812 users on 1.4.2 got the 1.5.0 js bundle overnight, how do i unship an update ota-update

Roll back first, understand second.

eas update:rollback on the affected branch, or republish the last known-good update onto it. Either way the fix ships as an update, which means it reaches clients as fast as they check - minutes, not a store review. Do that before you read the rest of this.

Then eas channel:view production to confirm what it actually points at now, and fix the mapping.

The structural fix is the runtime version. Both of your builds carried the same one, which is exactly why the 1.5.0 bundle was considered compatible with a 1.4.2 binary - the system checked, and you told it they matched. Move to the fingerprint policy, which hashes the native side so any native change automatically produces a different runtime version. Then a mismatched update is simply not offered to that client and this specific mistake becomes impossible rather than merely unlikely.

Once the fire is out, check whether the bad bundle wrote anything persistent. If 1.5.0 migrated the shape of anything in local storage, rolling the JS back leaves those 812 devices holding data from the future, and the old code will read it and do something interesting.

That is the part that shows up three days later as a mysterious crash on a small number of devices you cannot reproduce.

Operational rule that would have prevented this: channels are immutable once a build carrying them is in the wild. Never repoint one. If you need a different branch, cut a new channel and a new build. It feels heavy-handed until the first time it saves you, which was yesterday.

Gradual rollouts exist for this. Even 10% for one hour turns a 812-user incident into an 80-user annoyance you catch on a dashboard before anyone emails you.

Push a new store build with the correct bundle, that supersedes the update.

Small thing that pays off exactly now: pass --message on every update with what changed and why. When you are looking at a dashboard at 2am trying to work out which one was good, a list of commit hashes is not a list of information.