[finding] A Blocked-by: target parked in pm:on-hold or needs-user-decision can never fire the unlock — 6 measured cards waiting on a card that will not close #11219

Finding — recording only, not claimed. Filed under the maintainer direct-dispatch channel, this session, verbatim: 「然后 把 1–4 落成 finding 卡」. PM session session_0124Qg8rLvpXnQDwCmpKUmaJ.

The mechanism

The unlock predicate is "the Blocked-by: target CLOSED". pm:on-hold and needs-user-decision are, by definition, states a card sits in while open. A block naming such a target is therefore structurally indefinite: nothing in the machinery can ever fire it, and no predicate reports that.

It is not the same as H9 (a hold with no fireable Restart-when:). H9 audits the held card. This is about the waiting card, which looks perfectly well-formed — it has its machine-readable line, its target resolves, the target is open. Every check passes and the card still cannot move.

Measured, 2026-08-23

Two structural notes fall out of that table:

• objectos's entire blocked inventory (2 of 2) traces to one unanswered decision card, Complete protocol review and Q1-Q4 2026 action plan (Chinese + English) - Verified 2026-01-23 #68 — which is also the only item in the whole fleet's decision inbox. One ruling clears the repo.
• cloud#987 blocks two cards at once, and both (filters 参数格式与查询引擎不兼容导致筛选无效 #799, Add objectstack-hooks skill guide for third-party developers #1119) are eval-infrastructure cards that have been parked since July.

A third variant found in the same sweep: the stale chain

cloud#1395 → objectstack#10101, which is open — so the block looks live. But #10101 was itself an H19 finding on the same sweep: both of its own blockers had already closed. The block was real one level up and false two levels up, and a single-level predicate cannot see that.

Shape of the fix

A predicate over data the patrol already holds — it fetches the target's state to answer H19, so the target's labels are one field further:

an open pm:blocked card whose Blocked-by: target is open and carries pm:on-hold / needs-user-decision ⇒ name it.

Report-only, like everything else there. It does not say the block is wrong — sometimes waiting on a deferred card is exactly right. It says this block has no mechanism that will ever release it, which is the thing a human should see rather than discover in a hand sweep.

Optional second leg, cheap on the same data: flag when the target is itself pm:blocked (the chain case).

Re-check

# for each open pm:blocked card, resolve its Blocked-by target and read the target's labels