@alex-kuzmin-hg what was the impact to the node/network?

@poulok No visible side-effects or performance degradations observed. However, as this warning is new and may indicate unhealthy state, we need to get proper evaluation from engineers here

There are three different calls to ConcurrentArray.parallelTraverse(), where the exception was thrown:

• VirtualNodeCache.deletedLeaves() - a part of a flush, called on the lifecycle thread
• VirtualNodeCache.filterMutations() - a part of a flush, called on the lifecycle thread
• VirtualNodeCache.purge() - a part of node cache release, called on a thread in the cleaning thread pool

In the first two cases, the future returned from parallelTraverse() is then checked for exceptions using getAndRethrow(). All exceptions rethrown in this way on the lifecycle thread would be very visible in the logs (and most likely would kill or stuck the process). Therefore I assume the exception reported in this bug was thrown from purge(). It may cause a memory leak, but for the leak to result in an OOME, there must be a lot of such exceptions.

I don't know what could go wrong during purge(), so as the first step I'm going to improve logging in parallelTraverse(), so all underlying exceptions are properly propagated.