[OpenArm] Failure replay for builders integrators (intermediate)

We want practical OpenArm notes on failure replay.

When a run goes bad, what do you save immediately so the event can be reconstructed later without guesswork?

If you reply, include one failure slice that turned into a useful improvement after replay.

Pick your closest symptom to follow the right troubleshooting path.

Reproduce "[OpenArm] Failure replay for builders integrators (intermediate)" in a controlled loop, then compare baseline vs current measurements before applying partial fixes.

If residuals or drift fail your acceptance limits after warmup, switch to full recalibration/recovery workflow.

Use them as a checklist template first. Confirm interface names, fixture IDs, and safety conditions in your own cell before execution.