Seen whilst doing the stress test in bug 765. Whilst creating 32 sessions over a period of five seconds, the agent failed to respond to load info requests in a timely manner (10 seconds), failing four session requests. The machine had 8 cores, so it should have been able to handle this. It seems like something is blocking vsmagent an unsuitable long time. The numbers end up with less than 1 session per second and core, and we should be able to do better than that. (It also might be a good idea to stop getting loadinfo for every session request. We could cache it a few seconds at least).
