valgrind-plugin makes jenkins run out of memory under heavy use.
This is caused by too many (all) Report objects being retained in heap by Result objects.
Fix is about making such report objects on-demand or just-in-time, rather than always and too early (and as kept forever or so).
We at Ericsson have production (jenkins) instances that were recently hit by this bug, in terms of JVMs running out-of-memory.
We indeed clearly saw such many large report objects retaining the most of our heap as per profiling of the latter.