There's also the problem that if the surveillance is effective, they (the CIA et. al.) still can't necessarily run out and stop everything they know about without tipping their hand about how much they know, causing the adversaries to change their strategy and rendering any future surveillance ineffective.

So, they either need to wait for a truly catastrophic attack, or find a better way to foil attacks without making it look like they knew something they shouldn't have been able to know.

That's of course, assuming it does work, which is probably pretty charitable.