There is a rev limiter that is activated in neutral {and park?}.
If you disconnect the IAC, rev the engine, and it drops back to ~normal idle rpm immediately, but when you reconnect the IAC wires, rev the engine, and it take a while to drop back to idle, then the IAC is being controlled 'open' somewhat. The only difference is the computer signal getting to the IAC, so it must be a computer controlled activity.
The IAC is also used by the computer to act as a 'dashpot'. A dashpot will prevent the throttle from snapping closed, and killing the engine on an engine with a carburetor. The IAC does essentially the same thing by providing some 'throttle' as the butterfly closes, not choking the engine completely.
If the idle is rough, then you have spark, injector, or mechanical problems, but I think you are observing the normal idle instability if the throttle is snapped open and then closed quickly.
I understand that these engines also have a MAP sensor that 'knows' Manifold Absolute Pressure, or manifold vacuum. It may be that the vacuum reading is fed to the PCM to let it know how much to goose the IAC.
If you have a DVOM with a Hz reading, you can check the MAP's reporting capability based on the altitude above sea level and reported barometric pressure. The Hz will vary based on local pressure and a DVOM can report the Hz generated. I don't have the table of values, but if it is diddling with the Hz as the throttle is opened and closed, and the numbers raise and lower accordingly, it is likely to be ok. If not, it may be that the MAP is not telling the PCM that the engine throttle is closed, or is closed when it is not, so it in turn diddles with the IAC to keep the engine running.
tom