The bug is pretty insidious. I tried to find out how to trigger it, and after re-watching my screen recording I saw the pattern :

I think I experienced this a long time ago too but just moved on because it was another Resolve "bug," and used other tools like StoryToolKit AI for captioning, and I just used the integrated captioning tool here and there without too much customization, but I do more captioning now with Resolve itself.
Crossing the line separating the caption and the video tracks works... but then you could select a video clip by mistake.