lstar/aggregate: skip trivial 1-raw-sig + 0-children case

[ch4r10t33r]

Closes #747.

Summary

Extend the early-skip predicate in lstar/spec.py:aggregate():

# Skip cases where running the prover provides no consensus value:
#   - 0 raw + 0 children: nothing to aggregate.
#   - 0 raw + 1 child:    a lone child proof is already a valid proof.
#   - 1 raw + 0 children: a single-validator "aggregate" carries no
#                         information the raw gossip sig doesn't already
#                         carry — the sig is already on the gossip topic.
if not child_proofs and len(raw_entries) <= 1:
    continue
if not raw_entries and len(child_proofs) < 2:
    continue

Why

A 1-validator aggregate can't move ⅔ quorum and the raw XMSS sig is already on the per-subnet attestation_signatures gossip topic at sign time, so any peer aggregator can fold it in as a raw entry next round. The recursive STARK prover is roughly constant-cost in input size — building a 1-validator proof costs the same as a 32-validator one.

On the multi-client devnet, two zeam aggregators sat in this case for 100% of slots over 10 minutes, spending ~10.8 s of FFI per worker run on single-validator proofs and dropping ~50% of slot triggers as in_flight. Full numbers in blockblaz/zeam#907.

Correctness

• Wire format and verifier unchanged.
• The lone gossip sig stays in store.attestation_signatures because the existing bookkeeping only prunes sigs whose att_data produced an aggregate. A future round folds it in once another sig or a child shows up.
• Same end-state coverage as before, much cheaper to get there.

Tests

• New: test_aggregate_skips_single_gossip_sig_with_no_children — asserts no aggregation is produced for the 1-sig + 0-children case and that the sig survives in the store.
• Updated: test_multiple_attestation_data_grouped_separately — seeded with two raw sigs per att_data so each group is non-trivial. The test's intent (two distinct AttestationData → two distinct proofs) is preserved.

uv run pytest tests/lean_spec/forks/lstar/state tests/lean_spec/forks/lstar/forkchoice → 47 passed.

Related

• blockblaz/zeam#907 — devnet investigation
• blockblaz/zeam#908 — zeam-side fast path. Becomes spec-compliant once this lands.

[koko1123]

@ch4r10t33r I owe you an apology — I picked #747 up from the issue tracker and completely missed this PR, and ended up submitting a duplicate implementation as #1191. Closing mine in deference to yours: you wrote the issue, measured the devnet impact, and implemented it first.

Main has moved quite a bit since May (the spec/forks restructure, and fork behavior is now tested via consensus vectors under tests/consensus/ rather than pytest units), so in case it helps with the rebase, my branch is current against today's main and you should feel free to take anything from it, no attribution needed:

https://github.com/koko1123/leanSpec/tree/skip-trivial-single-signature-aggregation

A few notes from having effectively done that rebase, in case they save you time:

• aggregate() now lives in src/lean_spec/spec/forks/lstar/aggregation.py.
• The predicate can collapse to a single condition: len(raw_signatures) + len(child_proofs) < 2 is exactly equivalent to the two-condition form in the issue — both skip precisely the R + C <= 1 cases (verified by enumeration).
• On the harness dependency we both hit in build_signed_block_with_store: an alternative to the skip_trivial_inputs kwarg is to keep the spec signature unchanged and build the skipped single-signature proofs directly in the harness, mirroring how _sign_block already builds the 1-signature proposal component. That sidesteps any spec-purity question in review and keeps the 12 single-vote fork-choice scenarios' weights intact.
• The skip case can be covered with a fork-choice consensus vector: a single raw signature, a tick across interval 2 as aggregator, then assert the new-aggregate pool stays empty and the signature stays in the raw pool. There is a ready one on my branch (test_interval_2_aggregator_skips_single_raw_signature).

Happy to review the rebase or help any other way — this is worth landing either way for the zeam aggregators.