Yes, it’s a confusing experience for screenreader users. I hope the team can revisit this sometime soon.
FYI I ended up using Lity (Lity - Lightweight, accessible and responsive lightbox) for video lightboxes as it doesn’t strip the closed captions from YouTube embeds like the Cwicly one does (also an accessibility problem). It works well with a little custom CSS, but does mean you have to load jQuery on the front end. (see Closed Captions / Subtitles button missing in YouTube lightbox - #7 by Graphnic)