I disagree, this isn't the same bug as HHH-10286 Closed . In fact this bug is almost the opposite of it. The fix to HHH-10286 Closed caused HHH-10418 Awaiting Test Case which was really just the exposing of a new symptom of this bug (CCE on startup). That has since been fixed by splitting in to three maps along type lines (entity, collection, natural-id).

The central problem here is that CacheImpl uses maps keyed using the region name. This means two users of 2LC that access the same region get the same access strategy regardless of whether or not it matches their configuration. In my failing testcase a "read-write" configured entity sharing a region with a "nonstrict-read-write" entity gets it's cache consistency rules silently lowered to "nonstrict-read-write" without any indication to the user. In my eyes this is pretty broken.