One of our customers is running 3 separate services as activatable services under jini. Normally they run within a single shared JVM.
when they run the system they see random java crashes and their services get
restarted.

They've also seen signs that services restart in a similar manner on Solaris.

We've tried this with jini 1.1 and 1.2.1, we've tried jdk versions 1.3.1_02, and
1.3.1_03. We've tried running without ssl and without log rotation code, neither
of which appeared to make any difference.

From switching on RMIDs tracing it appears that the processes started by RMID
are exiting abruptly and being restarted.

Here is a stack trace using -XX:+ShowMessageBoxOnError flag.

oopDesc::mark() line 37 + 10 bytes
oopDesc::is_forwarded() line 254 + 8 bytes
Scavenge::scavenge_oop_with_check(oopDesc * * 0x03f5e9a4) line 114 + 8 bytes
InterpreterFrameClosure::offset_do(int 7) line 464 + 10 bytes
OopMapCacheEntry::iterate_oop(OffsetClosure * 0x02cffcc4) line 175
frame::oops_interpreted_do(void (oopDesc * *)* 0x0809c340
Scavenge::scavenge_oop_with_check(oopDesc * *), const RegisterMap *
0x02cffd24) line 606
frame::oops_do(void (oopDesc * *)* 0x0809c340
Scavenge::scavenge_oop_with_check(oopDesc * *), RegisterMap * 0x02cffd24)
line 742
JavaThread::oops_do(void (oopDesc * *)* 0x0809c340
Scavenge::scavenge_oop_with_check(oopDesc * *)) line 1801
Threads::oops_do(void (oopDesc * *)* 0x0809c340
Scavenge::scavenge_oop_with_check(oopDesc * *)) line 2621
Scavenge::invoke_at_safepoint(int 198, long 0, long & 0) line 555 + 12 bytes
VM_Scavenge::doit() line 108 + 21 bytes
VM_Operation::evaluate() line 35
VMThread::evaluate_operation(VM_Operation * 0x038df6a0) line 228
VMThread::loop() line 304
VMThread::run() line 167
_start(Thread * 0x0003f670) line 239
MSVCRT! 77c37fb8()
KERNEL32! 77e802ed(