Servicing occasionally results in `mesh child abnormal exit name="vmm" code=57005` during post-servicing AP startup

[smalis-msft]

The various servicing tests can very rarely result in an apparently crashing openvmm with exit code 57005 (0xDEAD). It is unknown where this exit code is coming from or what it means. Attempts to catch this in a debugger or dump have so far been unsuccessful.

Next step is probably to add process lifetime ETW tracing to the host.

[smalis-msft]

This hasn't been seen in 30 days, so closing for now.

[smalis-msft]

Seen again, but now copilot has some answers:

The 0xDEAD mystery is solved:
It is the Windows kernel's exit code when a Job object with JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE is closed — petri's Mesh holds exactly such a job (job.rs:33-60, lib.rs:400-440). It is set externally by the OS kernel when the petri test process drops its Mesh during cleanup 5 seconds after vtl0 start failed.