FWIW, the above M42 mosaic is the only one so far that's gone beyond 30 minutes of exposure time. I looked at the sub frames, and it looks like it did about 3 1/4 full passes over the frame in that 75 minutes of exposure time. .
Every other mosaic I've tried since gets to about 85% of one full frame's worth of coverage, and then the stacking error rate goes up and up until it's 100%.. For 1.4-1.5x zoom out frame this means about 22-23 minutes. I've tried to do the Horsehead Nebula under clear skies but same Bortle 6 conditions as M42 above, and all fail the same way.
Am running another tonight, and it started dropping every exposure right at 21:40, but I let it keep trying, and it failed completely for about 30 minutes, then stacked another 7 minutes of exposures, and now is failing at 100% rate again, stuck at 28:50 for the last 5 minutes. When it fails, it's usually in one of the corners, like upper left or lower right, and the impression I get is that it's struggling to align new rotated exposures over the top of existing high noise corners. What I leaned from the M42 sub frames was that it doesn't actually use a horizontal scanning pattern but it spirals out from the center which means the corners always show the least overlap and thus highest noise. If it's a dim target like IC434 and the corners are well outside the object itself then it's really just trying to align dim stars and noise. And it fails usually in the upper left corner, because that's the last part of the spiral out so the most rotation has occurred by the time it gets there. I think M42 made it further because it's so much bigger so it always found less dim details to align on. The attached imaged shows where it failed the first time.. Notice upper left corner is very noisy and incomplete.
And the next one shows where it's stuck.. again.. upper left corner.
Where I'm stuck now around around 29 minutes, it basically made another full pass around the frame reaching out beyond the borders, where you can see it's imaging outside the frame, but it can only align the stack where there's overlap in the saved part of the frame which is only about half the area. I wonder if that's part of why it struggles here.
