As this is a new camera and it was the first image with on a new target (IC434) that is containing a bright star (Alnitak), I cannot judge the gain of image visually. Well, I suppose the background might give it away. I'll look into that.
To the question when to actually refocus, I think if you realign before each step of a sequence, either do the alignment at the end of the last step, i.e. the last filter is still in focus, or before aligning with the new filter, and you can then use this focus for capturing. Although this might cause problems if a sequence is aborted/stopped and resumed at a later point and woudl have to be caught programmatically, i.e. 'always focus on teh first step'.
If align is disabled, you can just focus after a filter is changed or a certain time has passed, or temperature changes (is this implemented?) or HFR runs away.
Theoretically, I do not think that refocus should be mandatory for every step just because it's selected globally in the sequence tab. Consider bright cores of galaxies or nabulae with high dynamic range where one wants different expsure lengths for the same filter. Why refocus? It certainly won't hurt, and the time lost really is of little to no consequence, but making this conditional probably adds unnecessary complexity in both programming and usage. Just considering whether or not to, you just might want skip between certain steps.
Another thing is that I do not think of EKOS' focus functions as especially robust, so I personally would not want to have unnecesarry focusing runs that might or might not cause problems, potentially ruining an unattended imaging run
Sven