I don't really like this change. It makes thumbnails smaller even when text is visible - requires even taller strips to see the thumbnails.

Perhaps there could be small offset when there is no text, and this offset does not need to be on both sides. But I don't think there is need for this offset really.

Getting values from running job is quite bad idea - it can have outdated information.

Previously this used SEQ_STRIP_OFSTOP and SEQ_STRIP_OFSBOTTOM defines, which is fine. But (20 * U.dpi_fac * pixely) looks like it is just approximation of text height which is not very great. I didn't notice this during review, I would object to this line then.

Passing these y2 and y1 values is not as terrible really. Alternative would be to define text size similar to SEQ_STRIP_OFSBOTTOM and use this instead.

Even if you added #define SEQ_STRIP_TEXT_HEIGHT (20 * U.dpi_fac * pixely) it would be better than inlining the code directly. Then you need just way to know if text is drawn, this could be abstracted away too, and check can be reused where needed from context instead of passing boundaries. I don't have really strong opinion here, in practice you still need to mix these workflows, but I am not fan of passing 30 variables either.

Issue with adding utility functions to get these parameters is, that it could really use some general drawing code cleanup, because utility functions that look reasonable now may be completely unnecessary if it was structured better.

It looks like 1px offset being applied to height on top of thumbnail. Not quite sure why, but doesn't look necessary on first glance. Perhaps @Aditya Y Jeppu (quantimoney) does remember why/if this is necessary.

I don't see visible difference here

It's very weird to have this return variable then, it should be removed.

Could be simplified to 1 return value IMO, but that's not point of this patch.

I would declare const float thumb_height = y2 - y1; here, and remove argument from seq_get_thumb_image_dimensions as I mentioned in inline above.

Then pass it to sequencer_thumbnail_start_job_if_necessary too. Passing y2 - y1 is redundant math and not quite as readable.