Report checkpoint validation failures and support finite archive passes

[ElmoPA]

Report checkpoint validation failures and support finite archive passes

Bind W&B resume identity and report requeue signals correctly

Retire requests whose required dependencies can never pass

Document cluster source authority and durable restart workflow

Apply restart-safe artifact routing to the Action Flow stack

Verify recorded artifacts from separate CPU allocations

[ElmoPA]

Action Flow operational adoption of PR #32, preserving base to warmup to runtime-cleanup ancestry. Strict W&B resume identity and provenance, sparse telemetry health checks, signal/dependency handling, and mirror health. All three validation writers share restart-safe artifact paths; the smoke verifier handles matching training attempts and unambiguous evidence from separate CPU verification jobs. Tests passed: 82 operational cases plus two subtests; eight artifact cases; four final verifier cases (overlapping focused checks). No architecture, objective, loss schedule, dataset, or model configuration changes. No GPU run launched. Historical checkpoint sources are preserved.

[ElmoPA]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• Port three torus candidate objectives to conditional U-Socket graphs #39 : 2 dependent PRs (#40 , #47 )
• Report checkpoint validation failures and support finite archive passes #34 👈 (View in Graphite)
• Add reconstruction-only Action Flow warmup #26
• codex/usocket-action-flow-bc-20260905

This stack of pull requests is managed by Graphite. Learn more about stacking.