This was my thinking as well, it's safer to never delete data and simply null route requests to it. It's what I've done in the past when I was on their side of the ethernet cable.
I'm sure the 'why' this happened doesn't matter much to you now, but noticed in just searching for channel name "tranresearchtraining" google would auto-correct and search for "trans research training". Maybe that flagged the review algo to be extra sensitive to certain phrases, etc. Also, where you're dealing with multiple nationalities, wouldn't be surprised if it included non-native english speakers and/or words outside the standard english dicts... so their auto speech-to-txt may have taken creative liberties, and a mistranslation triggered other flags. So, basic edge case nightmare. Their automated support is the worst too. Probably worth trying to spread this story on twitter and linkedin as well to increase the chances of it reaching someone who might be able to help or create enough negative PR to get on the radar. good luck.
You could use https://www.nirsoft.net/utils/chrome_cache_view.html to extract your browser cache, then grep the whole thing for something like `UC[A-Za-z_-]{20,30}`. That'll probably find an overwhelming amount of junk, but it might also work too. (The idea is to search the cache for UCID matches in JSON dumps.)
If this seems interesting, the immediate priority would be to backup your cache folder immediately (%AppData%\Local\Google\Chrome\User Data\Default\Cache), then extract and search it at your leisure.
(This is all for Windows, but IIUC there are alternatives for Linux and macOS.)
The youtube channel in question.
https://www.youtube.com/user/tranresearchtraining/