🤖 fix(tasks): define in-process settlement authority across concurrent backends

[ThomasK33]

Problem

With two backends sharing a root, process A can retain its owned/settled attempt A while the persisted row names attempt B. The existing in-process classifier accepts its owned-pointer settlement as terminal-no-report; the new admission fence follows B. A native probe on PR #4308's candidate observed this divergence.

Evidence and scope

See #4308 (comment). Baseline 32834cca6 has the same owned-pointer classifier authority. The accepted recovery plan explicitly leaves concurrent backends outside the in-process-authority guarantee. This is a tracked limitation, not an implemented fix.

Acceptance

1. Define and test how in-process classification binds to the current persisted attempt under two concurrent backends, including refresh after awaited report reads and missing/unreadable rows.
2. Preserve the supported same-process recovery behavior or explicitly narrow it with tests and documentation.
3. Independently, G2's cross-process receipt/claim path MUST require evidence for the exact current attempt: settlement A must never authorize a receipt or claim for B. This follow-up does not waive that G2 gate.

Follow-up trigger: before claiming support for concurrent-backend in-process recovery; implement as a dedicated authority change, not additional mechanism in G1. This issue remains open independently of PR #4308.

---

Generated with xum • Model: coder:bedrock-mantle-us-west-2/openai.gpt-6-astra • Thinking: xhigh • Cost: $726.91

[linear-code]

MUX-163