Had my first observing run for this year last night, and it failed because autofocus couldn't finish after a filter change, and aborted the run. I had compiled a recent version of kstars and INDI from git 3 weeks ago.
(kstars-3.6.4-3558_g6d6a6f593.x86_64, libindi-2.0.1-2_g2a6b6d9e6.x86_64)

I analyzed the problem this morning, and found that the filter offsets were obviously applied in the wrong direction, so that after a change from blue to red, which needs an offset of ~100, the focus start position was off by ~200 units, which is too much for the L1P algorithm to find the proper focus

After I disabled the 'reverse motion' in the EAF INDI driver, offsets work again. I also verified that with it disabled the 'focus out' actually moves the focuser out. I'm very sure this was not the case last year, and I had to have reverse enabled for correct movement directions....
Git tells me there has been an SDK and driver update for the EAF since then. Did that change anything in that respect?

Has anyone seen similar issues?

Hi Peter,

I'm not aware of the issue you describe, although I see in the code that the driver supports this reverse motion functionality. I've been using a ZWO EAF for a couple of years.

The offsets and the setting of reverse motion would need to be set mutually consistently.

Hi John,

thanks for your comment. Yes, I'm aware the two settings need to be consistent, and they had been in the last years (I also use EAFs since at least 3 years). Computer and equipment had been switched off (and a few thousand km away from me) between December and April, and all I did was updating the OS and compiling the latest INDI and KStars ....

Guess I'll have a try and install an earlier version one of the coming days, and compare. If that works the same, it's probably my memory getting holes

Ha, I actually did something I didn't remember correctly. I had updated the firmware of my EAFs to 3.3.5. I thought that was much earlier, but the file in the Archive has a timestamp of Dec 17 2022, just before I shut down for the winter...

I just had a look at last years analyze logs from EKOS, and comparing the focus results I can see the problem starting between Dec 16 and Dec 19. Both images show the AF run from an automatic filter change, first one is blue->green, the second one green->blue. Those have an filter offset of 50 (my step size is 20).
The first one is nicely symmetric, indicating the start point of the run was close to focus. The second one is shifted by at least 3 steps, from symmetry even 6. In that case the offset was small enough that L1P did still find the proper focus. I had no change blue<->red in that session, else it likely would have failed, too...

Both days were run with the same version of INDI/KStars, so those are definitely not the culprit. Either it was a change in the firmware behavior, or it was me changing the setting. But it was on 'reverse on', and I doubt the default after a FW update would be 'on'. That I need it 'on' also matches my log entry of 2020 when I installed the EAF on the scope. Now I obviously need it 'off'.