The concurrency-interest mailing list mentioned a bug in the CHMv8 which could cause data loss, apparently a fix was applied already but I guess this doesn't magically appear in the Infinispan sources. I didn't create a JIRA as I haven't been able to look if the issue affects our copy, or how critical it is, but I guess someone should have a look at it as "data loss" doesn't sound too good; please could someone take this over and create one as needed as I'll be offline most of the next days. In this case I've been lucky to spot the discussion, but could we think of a better process to track the upstream code changes? Also it would be nice to catch these problems first, apparently others are doing more in-depth tests :-) Sanne