Just now I noticed my system had 8 plans aborted with the msg in the Scheduler log 'interrupted by acquisition process failure'. Last night ran flawless. So I RDP'ed into the obs computer and lo and behold ACP/Maxim is furiously trying to lock on a guide star, which is plain as day in the full guider window (>>50 SNR) and 'watch star' window, but Maxim keeps reporting 'Guide star fade'. So the thought came to mind, what is different today from yesterday?

Well this is the first day after winter where the temps have been really comfortable (>27C in Adelaide today, probably hotter up in Arkaroola). The guider is a SX Lodestar (no cooling), Maxim is v6.11, and I am using library darks (full calibration mode). So I turned off the guider image calibration and voila Maxim locked on. The bias/dark frames are from last May, ambient 8C at acquisition time, 22C currently, obviously the offset is temperature dependent enough to wreck the star finder in Maxim. Using the Maxim info window, I noticed the min and median pixel value in the aperture was 0, and the max a few thousand ADU. The bias must decrease with increasing temperature on this camera.

So the moral: the Maxim star finder is fragile to the background level being clipped (Maxim doesn't store negative pixel values, everything get truncated). Keep your uncooled guider bias/darks matched to the weather.
Now I can imagine why many others are baffled when they can see the star but the guider won't lock, it could be their bgd has no noise in it.