Perhaps by having drivers commit to a single platform, an hour at a time. At the end of the hour-long "shift" the driver then picks which platform to work for in the next hour, or chooses to go home.
Nice write-up! It would be interesting, however, to get more details on what types of errors were caught in dogfooding, which made it to production, what kind of hotfixes have had to be made in the past, etc...
It's nice to know what Slack does to mitigate bugs in releases, but it would also be useful to know what kinds of bugs each step catches and what bugs still slip through.
Does anyone know of the applicability of this to HomePod/Siri?