P. Bombas FWIW, folks are talking about several different issues in this thread, so sometimes the suggestions aren't relevant to the problem. If I understand yours, it's unstable azimuth, not altitude and thus has nothing to do with levelling. The problem is all the Seestar knows about Azimuth is from its electronic compass, which is not all that accurate (none of them are) and very easily disturbed by external ferrometallic materials possibly including things on your tripod itself. With the Seestar unmounted, I'd suggest you find a strong rare earth magnet and poke it around your tripod head and see if it sticks to any ferro magnetic material. Ideally everything should be aluminum, plastic and stainless (non magnetic steel) but if your magnet sticks to anything on or near the head it could be interfering with the Seestar's compass.
I just checked my big tripod (which I don't use with the Seestar) and it has some magnetic materials.
I use the stock Seestar tripod, with an aluminum and stainless, three screw leveler, and on top of that a fine screw adjustable alt/az wedge which is also aluminum and stainless, with a long aluminum rail (allows me to place the Seestar unit directly above the center of the small tripod even when tilted back 40 degrees), and a beefy aluminum 3/8" spacer on top of the rail.
Not a bit of it attracts a magnet.
Beyond that, I do find that during EQ polar alignment, if I have to move the AZ a lot (like 4-5 degrees) then it often is not very accurate so I'll get it close, then start over (back out of the menu and come back in and do a new alignment. Don't ever use the Refresh Deviation button) and zero in on it. I always set up my tripod in exactly the same spot and try to hand align it to the same direction visually each time, so usually my starting AZ isn't off by more than about 2 degrees and I can dial it in, in one shot.