I tested the last build tonight with 10 pixel dithering and it went very well. I acquired 18x600s and it lost the star only 2 times.
The star was lost when it was very close to the frame limit but even after failure the guiding continued and did not interrupt the sequence.
Sometimes the movements are always in the same direction and with a big pixel movement (10 pixels) it gets very close to the frame limit.
When it changes direction it goes fine again.

In fact, at least in my case (using Losmandy G-11), dithering ALWAYS moves in the same direction. You can easily check this by blinking the light frames that have been taking during a session. Usually I do use 10 pixel dithering and it doesn't take too many frames for the guide star to run into the subframe border. I could resort to using full frame guiding but that limits the frame rate quite a bit.

Would it be possible to "dither" (move) the subframe in lockstep with the guide star in order to have the guide star stay in the middle of the subframe?

I have a better question. How can a random dither always move in the same direction? That sounds like a bug to me. . .

Well, randomness is not truely random in computers, but still it would be odd to _always_ move in the same direction. I guess an invisible protection zone can be arbitrarily defined to bound the star to it to ensure it doesn't stray too far off.

This is from my most recent session. To me the dithering direction doesn't look random at all.

Another picture (taken weeks ago) using 2,8 pixel dithering resulting in "noise rain" also shows dithering always goes in one direction.

What strikes me is the factt that not only goes dithering in one and the same direction during a job. It appears to remain the same during the whole night.

Thanks, I just pushed a fix for the not-so-random dither and used different algorithm and add a bounded area. Please try the nightly build or wait for next release.

That's quite odd, since for me with 10 pixel dithering it does move in a different direction each time.

Backlash... could it be the cause? Even with 10 pixels movement?
My results are mixed. Sometimes it goes always in the same direction. And suddenly,then it change directions.
I have to test the new build.

Odd, indeed. But what can I say, it always moves in the same direction here as evidenced by the pictures I posted. I should have a chance to test the new build tonight. Backlash is not an issue with this mount.

Just ended tonight's session. Certainly useless to target nebulae at full moon but I wanted to test the new build. I updated nightlies just before I started the session. Here is the result of two 6-lights jobs. I don't see much improvement.