The first two days after a bad store push
A sequence for the forty-eight hours when crash mail arrives before lunch: timeline, cluster cut, rollback question, then sleep.
Field notes
Short pieces on crash clusters, freezes, and the awkward week after a store push — written for people who sit with the console, not for a general productivity audience.
Why a crash-free session number can lie
A quiet crash-free rate can hide a freeze on checkout. Stability is more than the absence of a native death.
What to bring to a stability review on Example Road
A packing list for the Petaling Jaya room: version codes, mapping files, device models, and the person who actually owns the suspected area.
ANR, hard crash, watchdog: three questions operations can ask
Duty managers do not need stack frames. They need to know whether the phone died, froze, or was killed for being slow, because the rollback story changes.
Reading a crash cluster without drowning in volume
How we pick three clusters when the console shows forty, using version filters and a paper timeline instead of sorting by raw count.