While investigating the latest EAC outage I found that there were too many large objects consuming old gen memory.

One of the offenders is the JiraChannelResponseHandler. See screenshot -

So 840 objects consuming an average of 70Kb each is not reasonable (for EAC anyway, especially in conjunction with some of our other heavy cache using plugins).

Regardless of any actual bug fix you might consider making could you please also get in touch with me and advise on what kind of tuning of caches etc I should be performing on EAC to avoid this much memory being consumed again.