While investigating issues on http://wiki.se.oracle.com/display/JPG/Lab+issues+and+improvements
I have noticed that almost all machines that have experienced problems with out of native memory have several processes running seemingly infinite loops with stacktraces such as:
        at vm.share.vmstresser.CompileAndDeoptimize$C.incv(CompileAndDeoptimize.java:19)
        at vm.share.vmstresser.CompileAndDeoptimize.incv(CompileAndDeoptimize.java:27)
        at vm.share.vmstresser.CompileAndDeoptimize.inc(CompileAndDeoptimize.java:32)
        at vm.share.vmstresser.CompileAndDeoptimize.access$200(CompileAndDeoptimize.java:3)
        at vm.share.vmstresser.CompileAndDeoptimize$1.run(CompileAndDeoptimize.java:44)
        at java.lang.Thread.run(Thread.java:722)

The hs_err tests must make sure that they clean up such stray processes (or make them terminate after a timeout).