Not sure if this is the best solution, but the old code is also incorrect.

The function unshimClientDB is used in migrations 32, 33, 35, and 39 it uses handleReduxMigrationFailure which modifies keyserverStore's cookie but cookie itself was moved there in migration 44 (code on the left is doing the same unfortunately).

We should probably do something like in the case of deprecatedUpdateClientDBThreadStoreThreadInfos where handleMigrationFailure is an optional param - but not sure if it's still worth it as it's touched old migrations.

Also, wondering what's the best pattern to avoid things like this in the future, I am skeptical about factoring out code from migrations to functions because it causes situations like this one.

Thanks @kamil... those concerns make sense. As this diff isn't regressing anything, I'll go ahead and land it, and then attempt to address the issues you brought up in a separate diff.

D11928 is the follow-up to address the issue @kamil brought up