Wait on quota, don't fail the shape

Quota as a scheduling signal rather than an error.

Porting the benchmark fleet to a third cloud, the launchers kept failing shapes on quota errors. The fix was to treat quota as a scheduling signal: wait and retry when CPUs or disk are exhausted, wait on expired tokens, keep polling when a describe call fails.

A tracker records which shapes are valid, which are waiting and which are done, so a campaign survives a stop and a restart.

Takeaway

What differed from the first cloud was not the benchmark. It was the assumption that capacity is there when you ask for it.

Client engagement, 2021 to 2026. Names, identifiers and internals are generalised.