I have noticed that when doing a JFR recording on a Java-program that has been alive on a computer that has slept for the night - most of the code cache is purged.

What I think is happening is that the code is considered cold - which it is, since the program hasn't been i use. The question is - why is the purge triggered by a JFR-recording and not happening otherwise?

We also need to investigate under what circumstances this happens - does the codecache need to be almost full for this to happen?