I've had a look at some successful and failed runs. The difference is again, purely timing-related. On the failed runs the pcache thread hasn't finished storing the cached entries by the time the next search starts. Inserting sleeps into the test script makes the problem go away. Also, configuring the cache DBs with dbnosync also makes the problem go away. I'll commit the latter change.
Pushed to RE24. Aaron, can you please re-test and confirm it is resolved in your setup? Thanks, Quanah -- Quanah Gibson-Mount Platform Architect Zimbra, Inc. -------------------- Zimbra :: the leader in open source messaging and collaboration