I’d usually trigger this error by unplugging the AC adapter and then picking suspend from GPM’s left click menu.

My first thought on seeing this was “What’s a policy timeout, and why is it not valid?” followed by “I don’t remember setting a policy timeout”. Looking at bug 492132 I found a pointer to the policy_suppression_timeout gconf value, whose description gives a bit more information.

Apparently the timeout is designed to ignore spurious messages from the hardware after a resume — you wouldn’t want to process a left over “suspend” message immediately after resuming from suspend after all. This does bring up a few questions though:

1. While ignoring “please suspend” messages shortly after performing a suspend makes sense, why ignore “please suspend” messages after an “on battery power message”?
2. While messages from the hardware might be spurious, surely picking an option from GPM’s menu is not. I guess such suspend requests are being mixed in with hardware suspend requests before the point where the policy timeout is checked.