I was hoping this wouldn't be necessary, but it seems that the possibility has increased now that we update the display name in the AB. If two SOAP requests fail because of outdated tokens for the same server, it's possible that the second one could fail miserably. That's because both times, the update request would be made with the original cipher secret. However, the response for the first request would overwrite this secret, and the second response would attempt decryption with this new secret instead of the original. Now, we queue up the callbacks if a token-update is already in progress. This results in a single update if there happens to be multiple failures at a time, and it stops this incorrect decryption problem. Fixes #8415.