So, I ran a test with just a few resources (~1000, but only about 100 should be returned) on the lookup resources and we still see a lot of deadline exceeded messages (client timeout is 5 seconds). It seems to build up, in the beginning, everything works like a charm but after some relations get deleted, we get this. We always use an old zookie to not break the cache and actually, it is totally fine if we get resources returned that have already been deleted. They would be filtered on the client side anyways. Any advice what we could do?