It's great to see you've noticed the difference ways to store fuzz targets between new/old experiments.

How about organize the logic to:

if FileSystem(fuzz_targets_dir).exists():
  # Check and return the trial count in new experiments.

# Check and return the trial count in old experiments.

There is a catch in counting the number of trials in the new experiment: Not all fuzz targets will be returned immediately at once. For example, we may have fuzz targets from trial 8 and 10 first, then from trial 2 and 5, and later the rest.
There are 2 solutions:

1. In most cases, we can assume run_all_experiments.py (which takes the trial count from --num-samples) is run under the same fresh environment as report generation. Hence we can add a line in run_all_experiments.py to record num-samples in an ENV VAR, and reuse it in report generation.
2. In rare cases where the ENV VAR is not set, we can assume the max trial ID so far is the trial count as a temp solution. E.g., if we have trial 01, 05, 10, then we assume 10 trials in total.

The workflow would be:

1. Check and use ENV VAR.
2. If not available, check fuzz_targets dir and count trials in new experiments.
3. If not available, check raw_targets dir and count trials in old experiments.
4. return a default value (0 or 1).

I reckon this is the minimum changes required to solve this, but please feel free to let me know if you can think of a more elegant/complete/sound solution : )

BTW, thanks for addressing this so quickly!