diff --git a/scripts/mma-backport-helper.py b/scripts/mma-backport-helper.py
new file mode 100644
index 000000000000..667c0b98219b
--- /dev/null
+++ b/scripts/mma-backport-helper.py
@@ -0,0 +1,620 @@
+#!/usr/bin/env python3
+"""
+mma-backport-helper.py
+
+Identifies MMA-related PRs on upstream/master since the release-26.2 branch
+point that have NOT yet been backported to upstream/release-26.2.
+
+The script works in three passes:
+  1. Path-based: finds first-parent (merge) commits on master that touch
+     known MMA-related directories.
+  2. Title-based: finds first-parent commits whose commit message matches
+     MMA-related keywords (catches PRs that only touch files outside the
+     canonical directories, e.g. server plumbing, metrics, etc.).
+  3. Author-based: finds PRs authored by MMA team members (catches PRs
+     that don't touch MMA directories or use MMA keywords, e.g. supporting
+     infrastructure changes).
+
+It then extracts PR numbers from each commit's subject line and subtracts
+the set of PRs already backported to release-26.2.
+
+Output is in topological merge order (oldest first), which is the order
+you'd want to backport in to minimize conflicts.
+
+Usage:
+  python3 scripts/mma-backport-helper.py [--fetch] [--upstream NAME]
+"""
+
+import argparse
+import re
+import subprocess
+import sys
+from dataclasses import dataclass, field
+from typing import Optional
+
+
+# ---------------------------------------------------------------------------
+# Configuration
+# ---------------------------------------------------------------------------
+
+# Core MMA packages and their test data.
+MMA_DIRS = [
+    "pkg/kv/kvserver/allocator/mmaprototype",
+    "pkg/kv/kvserver/allocator/allocatorimpl",
+    "pkg/kv/kvserver/mmaintegration",
+    "pkg/kv/kvserver/mma_store_rebalancer.go",
+    "pkg/kv/kvserver/asim",
+    # Supporting packages frequently modified alongside MMA.
+    "pkg/kv/kvserver/allocator/storepool",
+]
+
+# Perl-compatible regex for git --grep (matches against full commit message).
+# \b is needed because "mma" appears as a substring in words like "summary".
+TITLE_GREP_PATTERN = (
+    r"\bmma\b|mmaprototype|mmaintegration|storeliveness|\basim\b"
+    r"|allocator.*(mma|rebalanc)"
+    r"|storepool.*(mma|suspect|liveness)"
+)
+
+# MMA team members: GitHub username -> list of git author emails.
+# Used by the author-based pass (Pass 3) to find PRs by team members that
+# might not touch MMA directories or use MMA keywords in the title.
+MMA_TEAM: dict[str, list[str]] = {
+    "sumeerbhola": [
+        "sumeer@cockroachlabs.com",
+        "54990988+sumeerbhola@users.noreply.github.com",
+    ],
+    "tbg": [
+        "tobias.b.grieger@gmail.com",
+    ],
+    "wenyihu6": [
+        "wenyi@cockroachlabs.com",
+        "wenyi.hu@cockroachlabs.com",
+    ],
+    "angeladietz": [
+        "angela.dietz@cockroachlabs.com",
+    ],
+    "5hubh4m": [
+        "shubham.chaudhary@cockroachlabs.com",
+    ],
+}
+
+# Derived sets for quick lookups.
+MMA_TEAM_GITHUB_USERS: set[str] = set(MMA_TEAM.keys())
+MMA_TEAM_EMAILS: set[str] = {
+    email for emails in MMA_TEAM.values() for email in emails
+}
+
+# Python regex for filtering individual PR descriptions within multi-PR merges.
+MMA_PR_DESC_RE = re.compile(
+    r"\bmma\b|mmaprototype|mmaintegration|allocator|storepool"
+    r"|storeliveness|store.liveness|asim",
+    re.IGNORECASE,
+)
+
+# PRs to ignore (false positives or intentionally skipped).
+# Map from PR number to a short reason string.
+IGNORED_PRS: dict[int, str] = {
+    # Author-based matches not specific to MMA.
+    165450: "author-based: kvnemesis: add MVCC GC operation to KVNemesis",
+    166328: "author-based: replicastats: removed an unnecessary indirection in `replicaStatsRecord`",
+    166419: "author-based: roachtest: enable Go execution traces in kv/restart/nodes=12",
+    166421: "author-based: roachtest: enable Go execution traces in kv/splits",
+    166532: "author-based: investigate: analyze heap profiles on OOM failures",
+    166423: "author-based: roachtest: add Datadog Logs link to GitHub issues",
+    166560: "author-based: tsdump2duck: add CLI tool to convert tsdump.gob to DuckDB SQL",
+    167100: "author-based: engflow-artifacts: rewrite protobuf parsing with proper schema",
+    167211: "author-based: CLAUDE.md: improve PR description guidance",
+    166156: "author-based: intentresolver: chunk intent resolution to limit in-flight requests",
+    166698: "author-based: kvserver: introduce stageDestroyReplica and pendingReplicaDestruction",
+    166924: "author-based: server: use liveness cache for high-frequency scan callers",
+    167297: "author-based: intentresolver: expose intent budget constants as env vars",
+    166760: "author-based: kvcoord: fix a potential persistent failure loop in `DistSender`",
+    167289: "author-based: kvserver: stop treating split/merge trigger errors as replica corruption",
+    # Admission/AC SQL CPU token workstream (separate from MMA).
+    166745: "admission: plumb SQL CPU time token admission through SQLCPUProvider",
+    166792: "admission: integrate SQL CPU admission with CTT WorkQueue",
+    168231: "admission: add local token reservation to SQLCPUHandle",
+    168298: "admission: introduce CPU time token AC mode setting",
+    168383: "admission: add maxCPU burst qualification for resource groups",
+    168438: "admission: rename from tenantID to groupID",
+    168439: "admission: add resource group routing and per-group burst refill",
+    168829: "admission: add modeStrategy interface with serverlessStrategy",
+    169195: "admission: add composite groupKey type",
+    # SQL/UI/docs/CI unrelated to MMA.
+    166529: "sql: add WITH ZONE option to SHOW RANGES",
+    167927: "sql: rename zone_config_conformant to zone_config_needs_split",
+    167463: "docs: remove unused metrics.html",
+    168260: "ui/cluster-ui: fix empty cluster settings page",
+    168261: "ui/db-console: add 'Changed only' filter to cluster settings page",
+    168121: "roachtest: add CLAUDE.md with basic usage instructions",
+    168452: "ci: fix lint_test panic in affected-targets unit test runs",
+    168444: "ci: opt T-kv into auto-investigate for test failures",
+    # Roachtest/test infra not MMA-specific.
+    167120: "timeutil: fix timer pool vs synctest incompatibility",
+    167989: "roachtest: increase HTTP timeout in follower-reads test",
+    168156: "kvserver: deflake TestTenantCtx",
+    168243: "roachtest: bump post-test assertion timeout to 20m",
+    168248: "roachtest: extract artifact collection to artifactsutil",
+    167920: "roachtest: add WaitForFullReplication helper",
+    168047: "kvserver: isolate subtests in TestReplicaLatchingOptimisticEvaluation",
+    167616: "kvserver: skip TestStoreRangeSplitAndMergeWithGlobalReads under duress",
+    # Other unrelated kvserver/storage.
+    167295: "kvserver: remove ReplicaCorruptionError (follows ignored #167289)",
+    165577: "storage: add MultiEngineCompactionScheduler",
+    167853: "kvserver: add rate limiting for low-priority bulk read operations like TTL",
+    # Already backported but invisible to script — backport commits don't reference these PR numbers.
+    167051: "already backported via #167468 (release-26.2: asim changes)",
+    167110: "already backported via #167468 (release-26.2: asim changes)",
+    167897: "already backported (b5f05189765, d3ea3080d22, 5bae83c8aac on release-26.2; cherry-picks lack PR ref)",
+    167939: "already backported via #168475 (release-26.2: allocator: disable follow-the-workload when MMA is enabled)",
+    # Touches non-mma kvserver/load/allocator code, not mma-specific.
+    167861: "kvserver: improve store.capacity allocations",
+    # Test deflake, not mma-relevant.
+    167655: "storeliveness: skip flaky multi-store tests under duress",
+    # Roachtest-only scoring change, no production impact.
+    167696: "roachtest/perturbation: simplify scoring to impact ratios",
+    # Admission/AC SQL CPU token workstream (separate from MMA), batch 2.
+    169320: "admission: add rmStrategy for RM-mode burst refill",
+    169428: "admission: preserve non-tenant groupWeights across SetTenantWeights swap",
+    169708: "admission: retire tenant-weight cluster settings",
+    # Unrelated to MMA, batch 2.
+    169584: "sql: lower the priority of AUTO CREATE STATS job to BulkLowPri",
+    169336: "kvserver: plumb ScanStats to obtain pebble iterator statistics during request evaluation",
+    169658: "build/teamcity: upload phases.json artifacts for perturbation tests",
+    169659: "roachtest/perturbation: enforce 0.8x throughput floor",
+    # Master-and-onward only — explicit v26.3 default change, not for 26.2.
+    169430: "kvserver: enable multi-metric allocator by default in v26.3 (master+ only)",
+    # Master-and-onward only per intent.
+    169669: "allocatorimpl: retire kv.allocator.load_based_lease_rebalancing.enabled (master+ only)",
+    # Already backported via separate release PRs but invisible to script (cherry-picks lack original PR ref).
+    165719: "already backported via #169344 (release-26.2: mmaprototype: dynamically verbose logs for overloaded stores)",
+    169446: "already backported via #169590 (release-26.2: kvserver/allocator/mmaprototype: lock advisor methods)",
+    169230: "already backported via #169711 (release-26.2: mmaprototype: add structured snapshot of cluster state)",
+    169233: "already backported via #169734 (release-26.2: cli: collect mma allocator state in debug zip)",
+    169413: "already backported via #169742 (release-26.2: mmaprototype: count overloaded-but-blocked stores in MMA metrics)",
+    169741: "already backported via #169761 (release-26.2: mmaprototype: add captured_at to cluster state snapshot)",
+    169747: "already backported via #169876 (release-26.2: mmaprototype: materialize per-range constraint analysis in snapshot)",
+}
+
+# Regexes to extract PR numbers from merge commit subjects.
+# Bors merges look like "Merge #NNNNN #NNNNN ..." — all numbers are PRs.
+BORS_PR_NUM_RE = re.compile(r"#(\d+)")
+# Squash/trunk-io merges look like "title (#NNNNN)" — only the trailing
+# parenthesized number is the PR; other #refs in the title are issue numbers.
+SQUASH_PR_NUM_RE = re.compile(r"\(#(\d+)\)\s*$")
+
+# Format used to separate subject and body in a single git log call.
+# Using a delimiter that won't appear in commit messages.
+RECORD_SEP = "---<RECORD>---"
+FIELD_SEP = "---<FIELD>---"
+
+
+# ---------------------------------------------------------------------------
+# Data types
+# ---------------------------------------------------------------------------
+
+@dataclass
+class PRInfo:
+    number: int
+    title: str
+    merge_sha: str
+    source: str  # "path", "title", "author", or "both"
+    merge_pos: int = 0  # topological position (0 = oldest)
+
+
+# ---------------------------------------------------------------------------
+# Git helpers
+# ---------------------------------------------------------------------------
+
+def git(*args: str, check: bool = True) -> str:
+    """Run a git command and return stripped stdout."""
+    result = subprocess.run(
+        ["git"] + list(args),
+        capture_output=True,
+        text=True,
+        check=check,
+    )
+    return result.stdout.strip()
+
+
+def git_lines(*args: str) -> list[str]:
+    """Run a git command and return non-empty output lines."""
+    out = git(*args)
+    return [line for line in out.splitlines() if line]
+
+
+# ---------------------------------------------------------------------------
+# Core logic
+# ---------------------------------------------------------------------------
+
+def find_split_point(master: str, release: str) -> str:
+    return git("merge-base", master, release)
+
+
+def parse_merge_commit(raw: str) -> tuple[str, str, str]:
+    """Parse a record from our custom git log format into (sha, subject, body)."""
+    parts = raw.split(FIELD_SEP, 2)
+    sha = parts[0].strip()
+    subject = parts[1].strip() if len(parts) > 1 else ""
+    body = parts[2].strip() if len(parts) > 2 else ""
+    return sha, subject, body
+
+
+def extract_pr_numbers(subject: str) -> list[int]:
+    """Extract PR numbers from a merge commit subject.
+
+    Bors subjects ("Merge #NNN #NNN") list only PR numbers, so we extract all.
+    Squash/trunk-io subjects ("title (#NNN)") may contain issue refs in the
+    title text; only the trailing "(#NNN)" is the actual PR number.
+    """
+    if subject.startswith("Merge #"):
+        return [int(m) for m in BORS_PR_NUM_RE.findall(subject)]
+    m = SQUASH_PR_NUM_RE.search(subject)
+    return [int(m.group(1))] if m else []
+
+
+def extract_pr_title(body: str, pr_num: int, subject: str) -> str:
+    """Get a human-readable title for a PR from the merge commit body.
+
+    Bors merge bodies contain lines like:
+        159099: mma: plumbing changes for existing MMA metrics r=user a=user
+    We extract the title and strip reviewer/author annotations.
+    """
+    for line in body.splitlines():
+        if line.startswith(f"{pr_num}:"):
+            # Strip "r=... a=..." annotations.
+            title = re.sub(r"\s+r=.*", "", line)
+            # Strip the leading "NNNNN: " prefix.
+            title = re.sub(rf"^{pr_num}:\s*", "", title)
+            return title
+    # Fallback: for squash-merge style "title (#NNNNN)", strip the PR ref.
+    title = re.sub(r"\s*\(#\d+\)\s*$", "", subject)
+    return title
+
+
+def get_first_parent_commits(
+    master: str, split: str, extra_args: Optional[list[str]] = None
+) -> list[tuple[str, str, str]]:
+    """Get first-parent commits as (sha, subject, body) tuples.
+
+    Returns in git's default order (newest first).
+    """
+    fmt = f"{FIELD_SEP}".join(["%H", "%s", "%b"])
+    args = [
+        "log", "--first-parent",
+        f"--format={RECORD_SEP}{fmt}",
+        master, "--not", split,
+    ]
+    if extra_args:
+        args.extend(extra_args)
+    raw = git(*args)
+    if not raw:
+        return []
+    records = raw.split(RECORD_SEP)
+    results = []
+    for rec in records:
+        rec = rec.strip()
+        if not rec:
+            continue
+        results.append(parse_merge_commit(rec))
+    return results
+
+
+def find_path_based_prs(
+    master: str, split: str, dirs: list[str]
+) -> dict[int, PRInfo]:
+    """Pass 1: find PRs in merge commits that touch MMA directories."""
+    commits = get_first_parent_commits(master, split, ["--"] + dirs)
+    prs: dict[int, PRInfo] = {}
+    for sha, subject, body in commits:
+        pr_nums = extract_pr_numbers(subject)
+        is_multi = len(pr_nums) > 1
+        for num in pr_nums:
+            title = extract_pr_title(body, num, subject)
+            if is_multi and not MMA_PR_DESC_RE.search(title):
+                # Multi-PR merge: skip co-merged PRs that aren't MMA-related.
+                continue
+            prs[num] = PRInfo(
+                number=num,
+                title=title,
+                merge_sha=sha,
+                source="path",
+            )
+    return prs
+
+
+def find_title_based_prs(
+    master: str, split: str
+) -> dict[int, PRInfo]:
+    """Pass 2: find PRs whose commit message matches MMA keywords."""
+    commits = get_first_parent_commits(
+        master, split,
+        [f"--grep={TITLE_GREP_PATTERN}", "-i", "--perl-regexp"],
+    )
+    prs: dict[int, PRInfo] = {}
+    for sha, subject, body in commits:
+        pr_nums = extract_pr_numbers(subject)
+        is_multi = len(pr_nums) > 1
+        for num in pr_nums:
+            title = extract_pr_title(body, num, subject)
+            if is_multi and not MMA_PR_DESC_RE.search(title):
+                continue
+            prs[num] = PRInfo(
+                number=num,
+                title=title,
+                merge_sha=sha,
+                source="title",
+            )
+    return prs
+
+
+def find_author_based_prs(
+    master: str, split: str
+) -> dict[int, PRInfo]:
+    """Pass 3: find PRs authored by MMA team members.
+
+    Uses two sub-strategies:
+      A) Bors-style merges: grep for 'a=<github_username>' in commit body.
+      B) Non-bors merges (trunk-io, etc.): batch-check the author email of
+         each merged branch tip.
+    """
+    prs: dict[int, PRInfo] = {}
+    usernames = sorted(MMA_TEAM_GITHUB_USERS)
+
+    # --- Sub-pass A: bors merges with a=<username> in body ---
+    pattern = "|".join(rf"a={re.escape(u)}\b" for u in usernames)
+    commits = get_first_parent_commits(
+        master, split,
+        [f"--grep={pattern}", "--perl-regexp"],
+    )
+    bors_merge_shas: set[str] = set()
+    for sha, subject, body in commits:
+        bors_merge_shas.add(sha)
+        pr_nums = extract_pr_numbers(subject)
+        is_multi = len(pr_nums) > 1
+        for num in pr_nums:
+            title = extract_pr_title(body, num, subject)
+            if is_multi:
+                # Only include PRs whose line has a=<team_member>.
+                matched = False
+                for line in body.splitlines():
+                    if line.startswith(f"{num}:"):
+                        if any(f"a={u}" in line for u in usernames):
+                            matched = True
+                        break
+                if not matched:
+                    continue
+            prs[num] = PRInfo(
+                number=num, title=title, merge_sha=sha, source="author",
+            )
+
+    # --- Sub-pass B: non-bors merges – batch-check branch tip authors ---
+    # Get all first-parent merges with their parent SHAs.
+    lines = git_lines(
+        "log", "--first-parent", "--format=%H %P",
+        master, "--not", split,
+    )
+    # Build mapping: branch_tip_sha -> merge_sha (skip merges already handled).
+    tip_to_merge: dict[str, str] = {}
+    for line in lines:
+        parts = line.split()
+        if len(parts) < 3:
+            continue  # not a merge commit
+        merge_sha = parts[0]
+        if merge_sha in bors_merge_shas:
+            continue
+        for tip_sha in parts[2:]:
+            tip_to_merge[tip_sha] = merge_sha
+
+    if tip_to_merge:
+        # Batch-check author emails of branch tips in a single git call.
+        tip_shas = list(tip_to_merge.keys())
+        raw = git("log", "--no-walk", "--format=%H %ae", *tip_shas)
+        matched_merge_shas: set[str] = set()
+        for line in raw.splitlines():
+            line = line.strip()
+            if not line:
+                continue
+            parts = line.split(None, 1)
+            if len(parts) < 2:
+                continue
+            tip_sha, email = parts
+            if email in MMA_TEAM_EMAILS and tip_sha in tip_to_merge:
+                matched_merge_shas.add(tip_to_merge[tip_sha])
+
+        # Fetch details for the matched merge commits.
+        fmt = FIELD_SEP.join(["%H", "%s", "%b"])
+        for merge_sha in matched_merge_shas:
+            raw_detail = git("log", "-1", f"--format={fmt}", merge_sha)
+            sha, subject, body = parse_merge_commit(raw_detail)
+            for num in extract_pr_numbers(subject):
+                if num not in prs:
+                    title = extract_pr_title(body, num, subject)
+                    prs[num] = PRInfo(
+                        number=num, title=title, merge_sha=sha, source="author",
+                    )
+
+    return prs
+
+
+def get_backported_prs(release: str, split: str) -> set[int]:
+    """Find PR numbers already referenced on the release branch.
+
+    Backport commits reference original PR numbers in their branch name
+    or commit body (e.g. "backport-release-26.2-159099").
+    """
+    raw = git(
+        "log", "--first-parent", "--format=%s%n%b",
+        release, "--not", split,
+    )
+    # Match 5-6 digit numbers which are likely PR numbers.
+    return {int(m) for m in re.findall(r"\d{5,6}", raw)}
+
+
+def assign_merge_positions(
+    prs: dict[int, PRInfo], master: str, split: str
+) -> None:
+    """Assign topological merge positions to PRs (oldest = 0).
+
+    This determines the backport order: apply in ascending position order.
+    We only iterate over first-parent subjects (cheap string scan) and only
+    for the PR numbers we care about, so this is fast.
+    """
+    # Get all first-parent subjects in oldest-first order.
+    lines = git_lines(
+        "log", "--first-parent", "--format=%H %s",
+        "--reverse", master, "--not", split,
+    )
+    remaining = set(prs.keys())
+    for pos, line in enumerate(lines):
+        if not remaining:
+            break
+        # Quick check: only parse if any of our PR numbers appear in the line.
+        matched = set()
+        for num in remaining:
+            if f"#{num}" in line:
+                prs[num].merge_pos = pos
+                matched.add(num)
+        remaining -= matched
+
+
+# ---------------------------------------------------------------------------
+# Output
+# ---------------------------------------------------------------------------
+
+GITHUB_PR_URL = "https://github.com/cockroachdb/cockroach/pull/{}"
+
+
+def print_report(
+    missing: list[PRInfo],
+    ignored: list[tuple[PRInfo, str]],  # (pr, reason)
+    present: list[PRInfo],
+    total: int,
+) -> None:
+    print("=" * 64)
+    print("MMA Backport Candidates: master -> release-26.2")
+    print("=" * 64)
+    print()
+    print(f"Total MMA-related PRs on master since branch split: {total}")
+    print(f"Already backported: {len(present)}")
+    print(f"Ignored (skipped):  {len(ignored)}")
+    print(f"NOT YET backported: {len(missing)}")
+    print()
+
+    if missing:
+        print("-" * 64)
+        print("PRs NOT YET backported (in cherry-pick order, oldest first)")
+        print("-" * 64)
+        print()
+        print("  Legend: [P]=path-based  [T]=title-based  [A]=author-based  [B]=both(P+T)")
+        print()
+        for i, pr in enumerate(missing, 1):
+            tag = {"path": "P", "title": "T", "author": "A", "both": "B"}.get(pr.source, "?")
+            short_sha = pr.merge_sha[:11]
+            url = GITHUB_PR_URL.format(pr.number)
+            print(f"  {i:>2}. [{tag}] #{pr.number:<6}  {short_sha}  {pr.title}")
+            print(f"            {url}")
+        print()
+
+    if ignored:
+        print("-" * 64)
+        print("PRs ignored (add/remove in IGNORED_PRS dict)")
+        print("-" * 64)
+        for pr, reason in ignored:
+            short_sha = pr.merge_sha[:11]
+            print(f"  #{pr.number:<6}  {short_sha}  {pr.title}")
+            print(f"       reason: {reason}")
+        print()
+
+    if present:
+        print("-" * 64)
+        print("PRs already backported (for reference)")
+        print("-" * 64)
+        for pr in present:
+            short_sha = pr.merge_sha[:11]
+            print(f"  #{pr.number:<6}  {short_sha}  {pr.title}")
+        print()
+
+
+# ---------------------------------------------------------------------------
+# Main
+# ---------------------------------------------------------------------------
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Identify MMA-related PRs that need backporting to release-26.2.",
+    )
+    parser.add_argument(
+        "--fetch", action="store_true",
+        help="Run 'git fetch' before analysis.",
+    )
+    parser.add_argument(
+        "--upstream", default="origin",
+        help="Name of the upstream remote (default: origin).",
+    )
+    args = parser.parse_args()
+
+    master = f"{args.upstream}/master"
+    release = f"{args.upstream}/release-26.2"
+
+    if args.fetch:
+        print("Fetching upstream...", file=sys.stderr)
+        git("fetch", args.upstream, "master", "release-26.2")
+
+    # Find the branch split point.
+    split = find_split_point(master, release)
+    split_desc = git("log", "--oneline", "-1", split)
+    print(f"Branch split point: {split_desc}", file=sys.stderr)
+    print(file=sys.stderr)
+
+    # Pass 1: path-based discovery.
+    print("Pass 1: path-based (directories touched)...", file=sys.stderr)
+    path_prs = find_path_based_prs(master, split, MMA_DIRS)
+
+    # Pass 2: title-based discovery.
+    print("Pass 2: title-based (commit message keywords)...", file=sys.stderr)
+    title_prs = find_title_based_prs(master, split)
+
+    # Pass 3: author-based discovery.
+    print("Pass 3: author-based (MMA team members)...", file=sys.stderr)
+    author_prs = find_author_based_prs(master, split)
+
+    # Merge results: path takes precedence; mark "both" if found in multiple passes.
+    all_prs: dict[int, PRInfo] = {}
+    for num, info in path_prs.items():
+        all_prs[num] = info
+    for num, info in title_prs.items():
+        if num in all_prs:
+            all_prs[num].source = "both"
+        else:
+            all_prs[num] = info
+    for num, info in author_prs.items():
+        if num not in all_prs:
+            all_prs[num] = info
+
+    # Assign topological merge positions for ordering.
+    print("Building merge order index...", file=sys.stderr)
+    assign_merge_positions(all_prs, master, split)
+
+    # Determine which PRs are already backported.
+    backported = get_backported_prs(release, split)
+
+    # Split into missing, ignored, and present, sorted by merge position.
+    by_pos = sorted(all_prs.values(), key=lambda p: p.merge_pos)
+    missing = []
+    ignored = []
+    present = []
+    for p in by_pos:
+        if p.number in backported:
+            present.append(p)
+        elif p.number in IGNORED_PRS:
+            ignored.append((p, IGNORED_PRS[p.number]))
+        else:
+            missing.append(p)
+
+    print_report(missing, ignored, present, len(all_prs))
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/mma-backports.txt b/scripts/mma-backports.txt
new file mode 100644
index 000000000000..4c9913f4e266
--- /dev/null
+++ b/scripts/mma-backports.txt
@@ -0,0 +1,165 @@
+Branch split point: 44c386148e7 release: add update-workflow-branches command and TeamCity scripts (#159216)
+
+Pass 1: path-based (directories touched)...
+Pass 2: title-based (commit message keywords)...
+Pass 3: author-based (MMA team members)...
+Building merge order index...
+================================================================
+MMA Backport Candidates: master -> release-26.2
+================================================================
+
+Total MMA-related PRs on master since branch split: 76
+Already backported: 10
+Ignored (skipped):  66
+NOT YET backported: 0
+
+----------------------------------------------------------------
+PRs ignored (add/remove in IGNORED_PRS dict)
+----------------------------------------------------------------
+  #165450  4c123b411ff  kvnemesis: add MVCC GC operation to KVNemesis
+       reason: author-based: kvnemesis: add MVCC GC operation to KVNemesis
+  #166328  9fc2f395971  replicastats: removed an unnecessary indirection in `replicaStatsRecord`
+       reason: author-based: replicastats: removed an unnecessary indirection in `replicaStatsRecord`
+  #166419  7b2d2cff0dc  roachtest: enable Go execution traces in kv/restart/nodes=12
+       reason: author-based: roachtest: enable Go execution traces in kv/restart/nodes=12
+  #166421  630fc2d636f  roachtest: enable Go execution traces in kv/splits
+       reason: author-based: roachtest: enable Go execution traces in kv/splits
+  #166532  0e4c792e661  investigate: analyze heap profiles on OOM failures
+       reason: author-based: investigate: analyze heap profiles on OOM failures
+  #166423  7fdcb3c57fa  roachtest: add Datadog Logs link to GitHub issues
+       reason: author-based: roachtest: add Datadog Logs link to GitHub issues
+  #166560  3b786c5c307  tsdump2duck: add CLI tool to convert tsdump.gob to DuckDB SQL
+       reason: author-based: tsdump2duck: add CLI tool to convert tsdump.gob to DuckDB SQL
+  #166745  57c21590cc0  admission: plumb SQL CPU time token admission through SQLCPUProvider
+       reason: admission: plumb SQL CPU time token admission through SQLCPUProvider
+  #167051  3122e1b1742  asim: add sma-count and mma-only configs to skewed disk balance test
+       reason: already backported via #167468 (release-26.2: asim changes)
+  #166792  5f484ad0d8d  admission: integrate SQL CPU admission with CTT WorkQueue
+       reason: admission: integrate SQL CPU admission with CTT WorkQueue
+  #167110  72f5c45a5db  asim: document thrashing sources in skewed disk balance test
+       reason: already backported via #167468 (release-26.2: asim changes)
+  #167100  cc104061783  engflow-artifacts: rewrite protobuf parsing with proper schema
+       reason: author-based: engflow-artifacts: rewrite protobuf parsing with proper schema
+  #167211  043bbc66afc  CLAUDE.md: improve PR description guidance
+       reason: author-based: CLAUDE.md: improve PR description guidance
+  #166156  1d52a673164  intentresolver: chunk intent resolution to limit in-flight requests
+       reason: author-based: intentresolver: chunk intent resolution to limit in-flight requests
+  #166698  c2e82219150  kvserver: introduce stageDestroyReplica and pendingReplicaDestruction
+       reason: author-based: kvserver: introduce stageDestroyReplica and pendingReplicaDestruction
+  #166924  0c0f9c2bd16  server: use liveness cache for high-frequency scan callers
+       reason: author-based: server: use liveness cache for high-frequency scan callers
+  #167297  6394b41096b  intentresolver: expose intent budget constants as env vars
+       reason: author-based: intentresolver: expose intent budget constants as env vars
+  #166760  c279baa28e2  kvcoord: fix a potential persistent failure loop in `DistSender`
+       reason: author-based: kvcoord: fix a potential persistent failure loop in `DistSender`
+  #167289  461753032b1  kvserver: stop treating split/merge trigger errors as replica corruption
+       reason: author-based: kvserver: stop treating split/merge trigger errors as replica corruption
+  #167120  de52b5c9a99  timeutil: fix timer pool vs synctest incompatibility
+       reason: timeutil: fix timer pool vs synctest incompatibility
+  #167616  de13bf2c301  kvserver: skip TestStoreRangeSplitAndMergeWithGlobalReads under duress
+       reason: kvserver: skip TestStoreRangeSplitAndMergeWithGlobalReads under duress
+  #167655  a68374e2719  storeliveness: skip flaky multi-store tests under duress
+       reason: storeliveness: skip flaky multi-store tests under duress
+  #167463  57c74456a2b  docs: remove unused metrics.html
+       reason: docs: remove unused metrics.html
+  #166529  4a011da7d2e  sql: add WITH ZONE option to SHOW RANGES
+       reason: sql: add WITH ZONE option to SHOW RANGES
+  #167861  8b897c7f9dc  kvserver: improve store.capacity allocations
+       reason: kvserver: improve store.capacity allocations
+  #167927  42999aff1c1  sql: rename zone_config_conformant to zone_config_needs_split
+       reason: sql: rename zone_config_conformant to zone_config_needs_split
+  #167897  839d4a5a7cc  kvserverbase: add COCKROACH_DISABLE_MMA env var override
+       reason: already backported (b5f05189765, d3ea3080d22, 5bae83c8aac on release-26.2; cherry-picks lack PR ref)
+  #167295  f19780c8f0a  kvserver: remove ReplicaCorruptionError
+       reason: kvserver: remove ReplicaCorruptionError (follows ignored #167289)
+  #165577  a3ab149c214  storage: add MultiEngineCompactionScheduler
+       reason: storage: add MultiEngineCompactionScheduler
+  #167939  5ad75098f1b  allocator: disable follow-the-workload when MMA is enabled
+       reason: already backported via #168475 (release-26.2: allocator: disable follow-the-workload when MMA is enabled)
+  #168047  2058731abcf  kvserver: isolate subtests in TestReplicaLatchingOptimisticEvaluation…
+       reason: kvserver: isolate subtests in TestReplicaLatchingOptimisticEvaluation
+  #167989  a5419f09af4  roachtest: increase HTTP timeout in follower-reads test
+       reason: roachtest: increase HTTP timeout in follower-reads test
+  #168121  cd06503a6b5  roachtest: add CLAUDE.md with basic usage instructions
+       reason: roachtest: add CLAUDE.md with basic usage instructions
+  #168156  67790cf6e02  kvserver: deflake TestTenantCtx
+       reason: kvserver: deflake TestTenantCtx
+  #168243  2c3125fb1bc  roachtest: bump post-test assertion timeout to 20m
+       reason: roachtest: bump post-test assertion timeout to 20m
+  #168248  d721884544f  roachtest: extract artifact collection to artifactsutil
+       reason: roachtest: extract artifact collection to artifactsutil
+  #167920  e8766fc6d9e  roachtest: add WaitForFullReplication helper
+       reason: roachtest: add WaitForFullReplication helper
+  #168231  872d1c4c12c  admission: add local token reservation to SQLCPUHandle
+       reason: admission: add local token reservation to SQLCPUHandle
+  #168260  87c538de686  ui/cluster-ui: fix empty cluster settings page
+       reason: ui/cluster-ui: fix empty cluster settings page
+  #168298  eb8b75426e6  admission: introduce CPU time token AC mode setting
+       reason: admission: introduce CPU time token AC mode setting
+  #168452  8aca45b68b9  ci: fix lint_test panic in affected-targets unit test runs
+       reason: ci: fix lint_test panic in affected-targets unit test runs
+  #168444  17f510cfba0  ci: opt T-kv into auto-investigate for test failures
+       reason: ci: opt T-kv into auto-investigate for test failures
+  #168383  7b5db8f7a08  admission: add maxCPU burst qualification for resource groups
+       reason: admission: add maxCPU burst qualification for resource groups
+  #168261  79c8e71012c  ui/db-console: add "Changed only" filter to cluster settings page
+       reason: ui/db-console: add 'Changed only' filter to cluster settings page
+  #168438  8b6d3d2a14f  admission: rename from tenantID to groupID
+       reason: admission: rename from tenantID to groupID
+  #168439  6bf1e7400bb  admission: add resource group routing and per-group burst refill
+       reason: admission: add resource group routing and per-group burst refill
+  #168829  f4184f40c73  admission: add modeStrategy interface with serverlessStrategy
+       reason: admission: add modeStrategy interface with serverlessStrategy
+  #167696  51c340e4d27  roachtest/perturbation: simplify scoring to impact ratios
+       reason: roachtest/perturbation: simplify scoring to impact ratios
+  #167853  3a8907692f6  kvserver: add rate limiting for low-priority bulk read operations like TTL
+       reason: kvserver: add rate limiting for low-priority bulk read operations like TTL
+  #169195  0f45d1da541  admission: add composite groupKey type
+       reason: admission: add composite groupKey type
+  #165719  b7ceb33b651  mmaprototype: dynamically verbose logs for overloaded stores
+       reason: already backported via #169344 (release-26.2: mmaprototype: dynamically verbose logs for overloaded stores)
+  #169320  5da89382df2  admission: add rmStrategy for RM-mode burst refill
+       reason: admission: add rmStrategy for RM-mode burst refill
+  #169428  a7275e2a6f3  admission: preserve non-tenant groupWeights across SetTenantWeights swap
+       reason: admission: preserve non-tenant groupWeights across SetTenantWeights swap
+  #169446  9b6a9abf017  kvserver/allocator/mmaprototype: lock advisor methods
+       reason: already backported via #169590 (release-26.2: kvserver/allocator/mmaprototype: lock advisor methods)
+  #169584  92e3c33d1e1  sql: lower the priority of AUTO CREATE STATS job to BulkLowPri
+       reason: sql: lower the priority of AUTO CREATE STATS job to BulkLowPri
+  #169658  48bef1064d2  build/teamcity: upload phases.json artifacts for perturbation tests
+       reason: build/teamcity: upload phases.json artifacts for perturbation tests
+  #169659  8b8348c6f59  roachtest/perturbation: enforce 0.8x throughput floor
+       reason: roachtest/perturbation: enforce 0.8x throughput floor
+  #169336  3b12b3a7745  kvserver: plumb ScanStats to obtain pebble iterator statistics during request evaluation
+       reason: kvserver: plumb ScanStats to obtain pebble iterator statistics during request evaluation
+  #169230  ab29a85d378  mmaprototype: add structured snapshot of cluster state
+       reason: already backported via #169711 (release-26.2: mmaprototype: add structured snapshot of cluster state)
+  #169233  f7c159f74b7  cli: collect mma allocator state in debug zip
+       reason: already backported via #169734 (release-26.2: cli: collect mma allocator state in debug zip)
+  #169669  516a17d899f  allocatorimpl: retire kv.allocator.load_based_lease_rebalancing.enabled
+       reason: allocatorimpl: retire kv.allocator.load_based_lease_rebalancing.enabled (master+ only)
+  #169413  9a4e0558458  mmaprototype: count overloaded-but-blocked stores in MMA metrics
+       reason: already backported via #169742 (release-26.2: mmaprototype: count overloaded-but-blocked stores in MMA metrics)
+  #169741  7c25eaa9935  mmaprototype: add captured_at to cluster state snapshot
+       reason: already backported via #169761 (release-26.2: mmaprototype: add captured_at to cluster state snapshot)
+  #169747  cbc8baea715  mmaprototype: materialize per-range constraint analysis in snapshot
+       reason: already backported via #169876 (release-26.2: mmaprototype: materialize per-range constraint analysis in snapshot)
+  #169430  a7b7f3efaa1  kvserver: enable multi-metric allocator by default in v26.3
+       reason: kvserver: enable multi-metric allocator by default in v26.3 (master+ only)
+  #169708  3dc59b8f9bb  admission: retire tenant-weight cluster settings
+       reason: admission: retire tenant-weight cluster settings
+
+----------------------------------------------------------------
+PRs already backported (for reference)
+----------------------------------------------------------------
+  #166076  d71ec96bd45  storeliveness: fix flaky TestTransportClusterSettingToggle
+  #166148  d8f2867dd26  allocator/mmaprototype: add test for non-voter WB rebalancing
+  #166422  3fecb16a94a  roachtest: bump user_login.timeout in follower-reads insufficient-quorum
+  #162182  7ae2fd949c0  mmaprototype: balance CPU on utilization
+  #166306  4324c93284f  server: add cluster setting to disable GC assist
+  #166418  a9516dda64e  roachtest: fix acceptance/cli/node-status quorum loss
+  #166531  9e4eb90c76d  sql: fix TestReadCommittedReadTimestampNotSteppedOnCommit for write buffering
+  #166669  3aa8ac691bc  singleflight,stop: fix trace.StartRegion crash with oversized strings
+  #166658  70c832b9804  perturbation: rate-limit backfill pre-fill to avoid OOM on target node
+  #167307  bcc11cc1ac4  kvserver: fix TestReplicaRemovalDuringCPut hang on failure
+