Situation
Endpoint devices in the estate were failing in ways that looked like hardware lottery. Imaging and boot behaviour would not stay consistent.
Task
Find whether this was dead silicon or a repeatable firmware / imaging fault we could fix once for many units.
Action
Compared good and bad units, dug into UEFI and imaging assumptions, and drove a fix that did not depend on swapping boxes forever.
Result
The failure class stopped being “replace and pray.” (Stub — expand with exact model, root cause, and rollout.)