I find this pretty confusing, but I think this is going on:
1) I use a query to get an object A.
2) Object A has a collection of B's.
3) I use a different query to get a collection B's. This query is setup for caching.
4) I repeat step 3), expecting to get back the cached query results. But I don't. I took a look and think the problem is that when my query in step 4 executes, a Flush occurs. Now this Flush checks the "removeCollections" and for some reason the collection of B's on my A object is causing an invalidation to the query cache for the B's table. Well I have not changed the collection of B's on the A, so i don't see why the B's are getting invalidated. It seems just by virtue of there being a Persister on the collection, invalidation occurs.

Does this scenario seem plausible? I switched to using a readonly cache for the collection of B's on the A, and then got an exception. But nothing is modifying this collection so why would that occur?

Thanks,

JP

Apparently NHibernate thinks that something changed in that collection of B's or maybe it got detached from A somehow. How do property accessors in A look like? Also try to enable debug logging, there should be some hint in the logs as to what NHibernate thinks has changed.