I do recall experimenting as srmalloy did and ended up with the same conclusion. The NPCs almost always got an attack off. Whatever the delay was, it was short enough for the NPC to get an attack off.
These experiments are tricky to perform, because you're seeing many different effects simultaneously. You're seeing the effects of client-side prediction, network lag, some geometry interpolation, and differing server clocks. Furthermore, NPCs often had ranges that were longer than you might expect, which means at the instant you're in range the NPC might have already been in-range for a while. For example, Burst (AR) had a range of 90 feet for players - its range was intrinsically better than normal for such a power (typical range for such ranged attacks was 80 feet). However, the Council version of the same attack had a range of 100 feet - 10 feet more. So if you approached a Council NPC wielding this attack and it fired at you at the same time your queued attack fired at him, then in fact the Council NPC fired
significantly delayed from when they actually got within range - because he was in range before you were. And if you were using a more standard attack then the range difference was 20 feet, not 10 feet.
Because range had a number of complications, including just judging range and confirming the range of NPCs, the more accurate way to test the "reaction time" of NPCs was to test their ability to use melee attacks. Melee attacks were more limited in range and so the amount of time you were within the range of the attack was shorter, placing stricter limits on how fast the NPC had to react. And it was possible to engineer a test in which you were only within the range of the attack for less than a quarter of a second. In those circumstances, NPCs often did not get attacks off. In fact, a good way to test this was to use circular running motions to "orbit" the target, dipping into and out of melee range at high speed. In those circumstances, you could reduce NPC attack rates to dramatically lower than the recharge rate of their attacks. It was a known (to the devs and a very few players) exploitable bug in NPC behavior when facing players moving at extremely high speed.
A part of me wonders if effects like these (there were more than the one) caused the devs to be not all that sad that PvP suppression "bled into" PvE. The ways you could confuse NPCs with high velocity motion were numerous enough that I'm not convinced I actually know them all, because they were not commonly discussed.