From #26271: The current power virus implementation writes to the i2c device but has no logic enabled to flush the registers. This causes the test to hang on the second iteration when running in silicon. This change disables i2c writes if the target is kDeviceSilicon.
Backported to master in #28916
probably better to have logic to flush the registers. but OK with this for now. Maybe we could file an issue to track?
Originally posted by @timothytrippel in #28916 (review)