A problem has been observed with the BSBM EXPLORE+UPDATE benchmark where a query on a follower will sometimes fail if there is a concurrent update on the leader. This problem has been partly described at  (including a stack trace) and some work on the problem has been documented at [2,3]. It should be noted that the data is fine on the follower. If you halt the updates on the leader, you can read the data on the follower. The problem appears to be that the releaseTime is not being properly protected. The problem is demonstrated by
The root cause appears to be a failure to maintain the releaseTime correctly in the refactor to provide linear scaling in query for the followers .  is a related issue. The common concern is that the releaseTime must be updated correctly or active transactions will suddenly find themselves attempting to read on storage that has been either freed or reallocated.
The problem is present in r7106. Further work on this issue will be documented here.
 https://sourceforge.net/apps/trac/bigdata/ticket/530#comment:116 (HA Journal)
 https://sourceforge.net/apps/trac/bigdata/ticket/530#comment:117 (HA Journal)
 https://sourceforge.net/apps/trac/bigdata/ticket/530#comment:118 (HAJournal)
 https://sourceforge.net/apps/trac/bigdata/ticket/623 (HA TXS / TXS Bottleneck)
 https://sourceforge.net/apps/trac/bigdata/ticket/467 (IllegalStateException trying to access lexicon index using RWStore with recycling)