1. The dive that could not be aborted

James Cameron’s The Abyss returns one last time for this arc because the deep dives in the film do not allow easy retreat. If the suit fails, the diver dies before help can arrive. The desktop’s thermal problem is slower but similarly unforgiving once limits are crossed. A battery frozen solid or a processor thermally damaged cannot be fixed from the ground. This entry defines what success and failure look like for the test matrix defined in Entry 766.

2. Success criteria

The test as a whole passes if the autonomous thermal manager demonstrates:

  • Model accuracy: predicted temperatures remain within specified tolerances across all test conditions.
  • Thermal balance: all components stay within operating limits during hot-case and survival limits during cold-case conditions.
  • Correct heat trading: heat moves from hot cells to cold cells when the architecture allows, without causing runaway.
  • Timely protection: warning, caution, critical, and survival thresholds trigger the correct responses at the right times.
  • Graceful degradation: hot and cold emergency modes preserve hardware and commandability.
  • Clean recovery: when the environment improves or the fault clears, the system returns to nominal operation without overshoot.
  • Subsystem coordination: conflicts with power, ADCS, and payload schedules are resolved according to a documented policy.

A single row may be waived only if the failure is traced to a test artifact, not to flight software or hardware, and the waiver is documented.

3. Failure modes that stop the arc

The test arc fails, and the design must be fixed before flight, if any of the following occur:

  • Component damage: any flight-representative component exceeds its survival temperature limit during a recoverable test scenario.
  • False confidence: the autonomy reports healthy margins while the hardware is actually in a dangerous state.
  • Cascade failure: one thermal fault leads to a second, unrelated fault because the recovery sequence was wrong.
  • Silent failure: a thermal fault is not detected, not reported, or not logged.
  • Uncommanded mode: the autonomy enters a safe mode that ground cannot exit or diagnose.
  • Requirement mismatch: the autonomy behaves in a way that contradicts the documented thermal policy.
  • Model divergence: the thermal model cannot be correlated to test data within the required tolerance.

4. Partial success and the gray zone

Some results are not clean pass or fail. Examples include:

  • The autonomy survives the hot case but keeps temperatures higher than necessary, reducing component life.
  • The state estimate converges slowly, causing a brief but recoverable conservatism in scheduling.
  • A non-critical payload is throttled out of order because its thermal classification was wrong.

These are yellow flags. They do not stop the arc, but they require a documented mitigation: a model update, a procedure change, or an operational constraint. The arc cannot be closed until every yellow flag has an owner and a plan.

5. What this changes

  • Success is defined as trustworthy autonomy across modeling, balance, heat trading, protection, recovery, and coordination.
  • Failure is defined as any survival-limit violation, false confidence, cascade, silence, uncommanded mode, requirement mismatch, or uncorrelatable model.
  • Partial results are allowed but must be tracked to closure.
  • The next entry will close the thermal test arc.