[BUG]: Flaky driver initialization in CI #3183

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this bug and that I agree to the Code of Conduct

Type of Bug

Something else

Component

Infrastructure

Describe the bug

A crash-looping display-container process can race the subsequent device cycle and prevent the NVIDIA kernel driver from unloading

How to Reproduce

Not reproducible

Expected behavior

nvidia-smi -fdm succeeds always

Operating System

No response

nvidia-smi output

No response