Permit a 15-minute Polymarket Rust shadow gate with retained parity tail

[proerror77]

Parent

#326

What to build

Make the signed Polymarket release Gate eligible after a 900-second Rust shadow observation, while retaining the existing 601-second parity tail, real closed-segment preflight, immutable evidence, identity checks, OSS readback, and fail-closed rollback behavior. For this expedited Gate only, accept a policy-clean atomic Python health publication before shadow startup and continuous Python service identity during the Gate instead of waiting for a second Python full-catalog health publication that normally takes 35–58 minutes. The resulting evidence must accurately record that exception rather than reusing or relabeling a one-hour Gate.

Runtime control

• Target: ECS i-6we6afeqsvv8uo1ixmyo; only the five canonical Polymarket units; research/snapshot/Capture full-update Polymarket CLOB evidence for token microstructure research #235 excluded.
• Candidate: exact-main signed release rebuilt after this PR merges, with candidate, release-manifest, control-manifest, control-archive, and clean-ustar SHA bindings independently read back from OSS.
• Controller: /root is the sole production controller; all other work is read-only.
• Stop rule: stop the candidate shadow and retain Python if preflight, Rust health, Python identity, parity, OSS readback, immutable evidence, or the 1501-second Gate policy fails.
• Rollback: keep the pre-cutover Python unit control and revert the focused PR; if post-cutover verification fails, restore that verified Python control and stop Rust.

Acceptance criteria

• Production policy requires at least 900 seconds of observation and retains at least 601 seconds of parity tail.
• A focused counterexample test rejects evidence shorter than 900 seconds or with a shorter parity tail.
• Expedited legacy-Python evidence requires a policy-clean pre-start atomic health publication and continuous baseline identity, and records that post-start Python health completion was not required.
• A focused counterexample test rejects expedited evidence that omits pre-start health or fabricates post-start completion fields.
• All existing real-segment, Rust health, parity, OSS, and cutover preflight checks remain required.
• Gate evidence and its policy remain immutable and bind the exact candidate, control archive, source revision, and bundle.
• The change is independently buildable, reviewable, and rollbackable without changing collector, uploader, or research behavior.

Blocked by

#488

Out of scope

Changing uploader semantics, suppressing quote_collection_failure, changing collector health policy, research/snapshot/#235, or performing production cutover.