I had a moderately productive and successful test run last night (finally). I use PPA with my Astrotrac and the astrotrac wedge, head, and pier (the whole getup). I aligned it with the polar scope as best I could then took my vertical and horizontal shots. Unfortunately my local solves were not working quite as well as they had been with my previous data. I had to fiddle with the sigma value quite a bit... my default of 20 that had worked with my old data was sometimes too much and sometimes too little, but it always fairly close. If the solve failed with sigma 20 then I found I could get it to work at 15, and other times at 25. Unfortunately I think this part just requires trial and error as exposures are always going to be a little different (I had half the moon to contend with last night compared to last time when it was a new moon). BUT sometimes it would solve in 20 seconds, I'd check for improvement, move the axis literally just a smidge in one direction, take another shot, and then it would fail to solve despite using the same exposure settings! The field barely changed and it failed to solve when it had no problem on the shot before! Regardless, I was able to make multiple iterations to get it dialed in closer and closer, which is exactly what I needed. In the future I'm going to try to keep the histogram in the same spot as what has worked for previous sigma values.
Plus, my Nova solves weren't working because of ujson!!! I thought I had fixed that earlier but now I'm thinking maybe I only fixed it on my desktop during my testing and not my laptop. I even updated pip while out in the field last night (yay for cellphones and tethering) and tried updating ujson but it still didn't work! It errors out saying Microsoft Visual C++ 9.0 is required... I'll look into figuring that out later.
The best I could get was about 4 arcminutes. If my local solves were working faster without requiring me to adjust sigma every now and then I would have dialed it in further, but things were taking too long (my field laptop is ancient, it turns 10 years old this June but typically plate solves these for me within 30 seconds). Still, if it fails to solve then that means adjusting sigma up, adjusting sigma down, and sometimes still not getting a solve.
With this polar alignment I was able to take the following 4 minute exposure of M101 (very faint) from my red/white zone neighborhood with the Canon 6D and 400mm f/5.6. This is a 100% crop from the upper left 1/3 of the full frame:
A little trailing, but not bad, and certainly the best I've ever been able to achieve at 400mm and 4 minutes.