Big topic!
I’ve seen many people diffuse a stereo track beautifully with just a fader for each speaker (right panned to all right speakers, left panned to all left speakers, with volume control for each speaker on mixer faders). If the piece is already pretty active in the stereo field, then controlling front/back and up/down can be really fun and fascinating.
An additional approach could be to have EQ/Crossovers on the track(s) so that you can direct bass, treble, mids etc to different points in space. Audium in SF is 4 channels with subs in the floor, woofers at ear level and tweeters up high suspended from the ceiling. This spatializes the spectral changes. I’ve also played 8 channel shows with highs, mids and lows moving separately and it can be really wonderful.
If I’m composing a new piece I’ll do a combo of ambisonics and more direct panning - one issue with ambisonics is that (in my opinion) it does middle- and back-ground stuff really well, but I can rarely get something to feel really close, like inside your face, with it. So I use “regular” panning in Max for the foreground elements.
(Note: the ICST toolkit in max has doppler, swarming, and all kinds of stuff to make movement of sound more psychoacoustically realistic. Highly recommended if you have the time.)
One thing I strongly discourage is swirling. I personally find it cheesy when sounds spin around the listeners; I much prefer to think about them as events in a spatial environment that will move but not in a gimmicky way. Just my 2c.