apology for the delay. Can you file an issue and describe how to
reproduce the memory grow? We run really large test as part of GlassFish
v3 integration and so far we haven't measured such peak. So I want to
investigate how I can reproduce the issue locally and see what can be
improved.