I have noticed that when Autofocus fails, I get the message "Autofocus failed, moving back to initial focus position XXX". In my experience that position "XXX" is always a worse place to start again than if EKOS had started at the BEST focus position that it found. For example:

I start an autofocus run at position 5400 and it loops through and tries a final position of 3800, which fails as my focus criteria was not met,
So EKOS moves back to the start position of 5400. In all my experiences, this is ALWAYS a worse place to start than if it had just re-run autofocus from 3800.
What I do when I see it fail, is I abort it. Move the focuser to the best achieved position from the failed run, and begin from there. That always works.
Going back to the start position that failed is never the best answer in my set up (C11 and Moonlight Litecrawler).

Hi Peter,

When it fails after starting from 5400 I'm guessing that's because it fails the R2 Limit you set?

So, the main question would be why it's starting focus at 5400 when the solution is near 3800? I've been having some conversations with someone else who has problems with the first Autofocus run of the night. In his case the last Autofocus run of the previous session was at a cooler temperature and further "in". So his first run of the night starts too far "out" and sometimes works / sometimes not. Maybe your issue is something similar ?

If so, have you tried using the "Adapt Start Pos" option? This is part of the Adaptive Focus work and can be used either with or without using Adaptive Focus to manage focus position between subs. The idea is that Focus adjusts the start position of an Autofocus run based on filter and environmental differences (temperature and altitude) between when the last Autofocus run completed successfully and now.

So, for example, lets say your last Autofocus run was at 6000 and 10C. Currently, its 15C and you have configured your equipment's temperature parameter to be -100 ticks / degree C (in Filter Settings). If you have Adapt Start Pos checked, Focus will take its current position (6000) and apply the temperature offset 5 x -100, and start Autofocus at 6000 - 500 = 5500.

I've made a bunch of assumptions around your issue that may / may not be valid. If your issue is something different can you explain why Autofocus is starting a long way from where the actual solution is please.

I'm facing the exact same issue lately!

On a single night, I opened up the observatory 45 minutes before my session and the first autofocus was better.

But yeah, I can see donut shape stars at the beginning.

Just opened up my observatory and captured a frame to calculate HFR. Look at the value and those stars! Last night the HFR was 1.18. So, as far as I can understand, waiting to accumulate the temperature is a must.
In my case scenario, approximately every 10 minutes , HFR drops 0.25.