Since going to 7.5 we've been having an issue with image deployments occaisonally timing out on the Boot To Production task. By turning on additional logging on the task server we have been able to track this down to the new 'Registration Period' security feature. The default registration period is five minutes. On slower hardware, it often takes longer than five minutes to complete mini-setup after image deployment. What happens when this occurs is that the task server a) stops trying to tickle the production agent to resume running tasks and b) resets the task queue for the machine. This causes no further tasks from your deploy job to run and the job fails 60 minutes later (by default) when the Boot To Production task times out with the error: 'The task timed out on the Task Server while waiting for the agent to pick it up.' The solution to this issue is, of course, to increase the Registration Period on the Boot To task to a more suitable period (we chose 30 minutes).
Personally I would classify this as a bug, since it seems to me that when the task queue is cleared the current task should immediately fail with an error stating the registration period lapsed. This would point people directly toward the cause of the issue rather than making them wonder why their jobs are mysteriously timing out even though they can run new tasks on the machine successfully during the timeout period.