Just remember that a "Quick Delivery" relies on the un-reliable task engine. It sounds convenient, but it's better to set a time stamp, and then send out an update configuration request instead and simply let the policies handle it.
Even assuming that Task is working at 100% (that is, statuses are repoting up, things don't time out, etc, which right now for you they are not), Task is still less reliable. It has the advantage of "doing things now" - or rather, of TRYING to do things now, but it has definate weaknesses which is why it should be used very sparingly.
For instance, the task history stores 200K rows and then wipes. That's right - wipes the extra rows. Task history GONE. No alternate store, nothing. If a task expires before completion, it always shows failed, no matter if it actually succeeded or not.
Contrast this to policies that don't time out until you tell them to, can run even when the computer is disconnected, will report up success hours later if need be (e.g. if it ran when disconnected), keep their data in the logs for at least up to or exceeding 1M rows, have built in reports that gather succes rates that are very mature...
Using tasks for almost anything outside of DS is futile. It should be used as emergencies only (which I'll grant this might have been).. 0-day exploits for instance you might not want to be patient for policies for. :P
Quick delivery and tasks are very seductive, just not very efficient. Stick with the slower policy method and plan just a bit ahead and be paid off well over the long haul.