You can customize which format to download by doing first:
yt-dlp -F <url>
Check what numbers are there, and then download the formats you want.
You can customize both the video format & rate, and same for sound, by using
yt-dlp #+# <url>
Where the first hash is the id of the video stream, and the 2nd hash is the id of the audio id.
Take into account that you'd probably need also ffmpeg available on the command line for this; that if you "mix" containers (Say, a webm video and an m4u audio stream) you-ll end up with an .mkv container; and finally, that all of the downloaded streams are themselves containers (.m4u, mp4 & webm).
He's complaining about getting an .mkv in the first place, so this seems unlikely to solve his problem.
My guess would be that yt-dlp is giving you whatever file format it finds, with no editorializing. The default is just to go with "best quality" available. So "Is there a reason for this?" is likely a better question for whoever maintains the site(s) you're downloading from.
When using yt-dlp, I get a .webm file, and when asking for audio only using `-x`, I get an .opus file. Is there any reason for this?