FULL PRODUCT VERSION :
A DESCRIPTION OF THE PROBLEM :
As reported in bug 6457123, timed waits for instances of java.util.concurrent classes can leak memory. As it is not going to be fixed, this bug should be documented in the Javadoc for the API of Lock, Queue, CountDownLatch, and so forth. Even a short note mentioning the bug number 6457123 would be helpful.
If the problem were documented at the API level, it would likely save thousands - perhaps tens of thousands - of hours of debug time over the coming years.
I spent about twenty hours figuring out that the genuinely odd behaviour we were seeing in the production app was due to a memory leak.
This bug can be reproduced always.