Non-Reproducible Test Vectors due to Fresh aux_rand

[siv2r]

I'm implementing test vectors for ChillDKG. To ensure reproducibility, I use fixed values for random and hostseckey inputs, so every run of the test vector generation script yields the same vectors.

However, two internal functions, certeq_participant_step and pop_prove, currently call schnorr_sign using a freshly generated random 32-byte value for the aux_rand parameter. This fresh randomness prevents the consistent reproduction of test vectors.

bip-frost-dkg/python/chilldkg_ref/chilldkg.py

Lines 81 to 83 in 1e34161

	def certeq_participant_step(hostseckey: bytes, idx: int, x: bytes) -> bytes:
	msg = certeq_message(x, idx)
	return schnorr_sign(msg, hostseckey, aux_rand=random_bytes(32))

bip-frost-dkg/python/chilldkg_ref/simplpedpop.py

Lines 38 to 41 in 1e34161

	def pop_prove(seckey: bytes, idx: int) -> Pop:
	sig = schnorr_sign(
	pop_msg(idx), seckey, aux_rand=random_bytes(32), tag_prefix=POP_MSG_TAG
	)

Specifically:

• participant_step1 produces varying outputs (pmsg1) between runs because the Pop component generated by pop_prove differs each time, even with fixed inputs.
• Similarly, participant_step2 outputs vary because of the randomness introduced by certeq_participant_step.

These output variations propagate, affecting other participant and coordinator APIs relying on participant_step1 and participant_step2.

Is it a good idea to replace the fresh randomness with derived values (via a tagged hash) to achieve reproducibility?