The majority of my imaging in EKOS has been by setting up a camera sequence, performing a focus, plate solve, start guiding, then run the sequence. This has been fairly successful until I lose the guide star for some reason (either clouds, or trees get in the way). Then it never attempts to go to the object again, plate solve, and resume.

So I decided to start using the scheduler. Picked my target, and I encounter the exact same issues repeatedly over the night for various reasons. Either a wind gust sets it off a tiny bit, or clouds interrupt guiding, but in either case, the original guide start is not reacquired, a new goto with plate solve isn't done, and over time, I'm no longer imaging the original object (or it's no longer centered). Also, for some reason when using the scheduler, the meridian flip isn't happening. I've had to force it to flip but sending a goto command.

Aside from the above issues. The scheduler state really gets screwed up when you need to interrupt it. If for instance, I need to pause the scheduler, and submit a goto command to get back on the object, then resume the scheduler, it will mess up the sequence. My sequence is Lx5, Rx3, Gx3, Bx3, and repeat until dawn. So, when I pause the scheduler because it's off the target and submit a goto command, then resume the scheduler. It might have completed L and R filters. For the rest of the night, it will only capture G and B, because it thinks L and are are finished, and it doesn't loop back to the beginning of the sequence.

What I've had to do to fix these issues, is to quit out of EKOS. Relaunch it, and restart the scheduler from scratch. I also have to rename the folder it was saving into, so that it will create a new folder. Some times it seems there is some residual hold over issues if I keep the first batch of files in the directory. I could be wrong about that last part, but I think at least once or twice it caused an issue.

Which version of Ekos are you using? From my experience, the scheduler worked best in version 2.9.8 with the caveats I described earlier
indilib.org/forum/wish-list/4249-solid-e...i-version.html#32295
To fix the problems you describe, it might be helpful to have a feature that allows the user to force plate solving and recentering after a certain stated number of frames. So if the weather looks iffy, you could opt for plate solving after every 5, 10, 15 frames, etc, for instance.
Meridian flip is no problem using EqMod and an Atlas Pro mount with version 2.9.8. Works very reliably.
Jo

I'm using 2.9.8 on the Mac. I get frames exactly like you did your example after guiding is lost. Rather than force a new plate solve at an interval, I think the plate solve should happen after you lose the guide star. So if it's clouds, or a wind gust, and the guide star jumps, or disappears, this should trigger a plate solve, and imaging should not resume until the plate solve is successful and guiding has resumed.

The meridian flips are working for me in 2.9.8 when I don't use the scheduler and only use the sequencer. But when I use the scheduler, I go way past the meridian then have to force a goto to get the mount to flip. (using an AVX).

I've had two nights like this that were very unsuccessful. I restarted the computer tonight (not done on previous nights), and it's run flawlessly on a new target, running the same sequence. It even did the meridian flip with no issues. Don't know what caused the prior issues. The equipment wasn't touched from last night, and so far tonight it's run perfectly. I don't have a single lost frame in about 4 hours.

I've not run into any guiding issues with clouds, trees, wind, etc. So maybe that was the problem.

2.9.8 is actually riddled with scheduling problems that Eric is working so hard to fix for 3.0.0 release. Now 2.9.8 is still better than 2.9.7 when it was quite broke. We are very close to releasing 3.0.0 once Eric gives the go for the scheduler.

I just pushed an update that should help mitigate this issue. It watched for Pier Side property and if it changes then it switches the DEC swap so no recalibration is necessary. I have to test it once the weather clears up.

I support the change. It is equivalent to the "reverse DEC after flip" in PHD2.

-Eric