At scale some AWs to do not enter in complete state

[asm582]

Describe the Bug

At scale, some AWs do not enter into a complete state due to the fact that the informer and etcd do not agree.

Codeflare Stack Component Versions

Please specify the component versions in which you have encountered this bug.

Codeflare SDK:
MCAD:

Steps to Reproduce the Bug

Fire 1K AWs with very short jobs (10 seconds) and wait for completion of all 1K AWs

What Have You Already Tried to Debug the Issue?

I have run scale tests to reproduce the issue

Expected Behavior

All AWs should be completed.

Screenshots, Console Output, Logs, etc.

NA

Affected Releases

Current 1.35.0 release and main branch

Additional Context

NA

Add any other information you think might be useful here.