Comment 9 for bug 1421455

So looking at the powerd code, the only thing that does anything *real* between "exiting suspend" and "Emitting signal for transition to state ACTIVE" is a call to libsuspend_exit_suspend().

Which -- depending on the backend -- either makes one syscall and returns, or makes one syscall and waits for the fb state to be awake. Since the latter is the default backend and we're seeing delays, I'm guessing that's the code path we're hitting.

Why do we need to block on fb state there? (honest question, I don't understand the libsuspend code very well)