wvreeven Is there some kind of guide log on the ASIAir that I can inspect nd perhaps share here so you can see for yourself?
The PHD2 guide logs in ASIAIR keeps appearing and disappearing depending on the ASIAIR firmware release. If it is there, it should be in a Logs folder. (There must be an internal fight between people who want to give users more visibility of their instrumentation, and the idiots that want to make it "Simple as 1,2,3")
PHD2 is just a feedback loop -- measure the error, send correction pulse... rinse and repeat.
Check "Bode plots" for the classical treatment of feedback stability. I.e., what causes oscillation in a feedback loop. With discrete time loops, latency is of utmost importance since it is the analog of the phase term of a Bode plot. I.e., it can turn a phase of -180º (negative feedback) into a phase of +180º.
If there is a lot of latency in your system (i.e., takes a long time between measuring an error, sending the correction pulse and have the mount move), and if you cannot reduce the latency, your options are to turn loop gain down (this is the aggressiveness value in PHD2), increase the sampling frequency (remember Shannon-Nyquist?).
If turning aggressiveness down to 10% still produce oscillations, try changing the exposure time to 1 second or 0.5 second. With multi-star tracking, the variance of the error of "seeing" of a guide star is reduce by a factor of 2 for each doubling of stars used. That makes the need to use long exposure time to reduce seeing error pretty much disappear by the time you use 8 stars or more. So, just shorten the exposure (and ASIAIR will increase the frame rate it is sampling the guide camera -- ever since v1.5, ASIAIR uses synchronous ("video") mode for guiding). This should reduce the feedback latency, and reduce the ptntail of oscillatory effect.
Remember that in the final analysis, if your mount's motor has enough backlash (of the order of 0.2 arc second), it will also introduce what looks like latency.
I haven't seen any ill effect in autoguiding since v1.7 came out. In fact, it might be better than v1.6, when multi-star guiding was first tested on the guinea pigs (paying customers) -- but it could just the variability of clouds, Moonlight, winds, than actual ASIAIR code changes. I was seeing between 0.35" and 0.45" total RMS last night with the RainbowAstro RST-135, whose native periodic error is around 70 arc seconds peak-to-peak. i.e., starting around 25 arc second RMS error just on one axis).
I use 0.5 second exposure times with a Borg 55FL as the guide scope, a monochrome ASI290 (USB 3; not mini) camera, and a 685nm IR pass filter (I do near-IR autoguiding). The feedback latency is probably as low as you can get with the ASIAIR (the harmonic drive gearin the RST-135 has virtually no backlash).
EDIT2: removed comment about log missing in v1.7.