The cache now uses the site identifier as part of the hash prefixed to all keys.
Within its handling it attempts to load the site identifier from the cache config. If its not there then the cache config tries to load it from the database config.
The issue arises that if you write a cache definition that is used before the database cache then you can find yourself in a situation where the cache ends up in an infinite loop while trying to initialise.
The cause of this is that the cache initialisation is trying to use the cache API. That's a no-no the cache API cannot use caches or we risk the infinite initalisation issue.

There's a simple fix for this luckily.