A reported critical warning is masked to stop repeat notifications, and
stays masked until the next controller reset, so a second over-temperature
event goes unreported. Poll the health log while any warning is masked
and re-enable each one when it clears (Base 2.4 Annex B.4).
Asynchronous Event Configuration and Critical Warning: NVM Express Base
Specification 2.4, sections 5.2.30.1.5 and 5.2.13.1.3.