You're misunderstanding the core concept here. I'm not saying mRNA spontaneously becomes healthy, fully-functional non-coding RNA units. I'm saying if you force a huge number of random RNA sequences in there some of them might be just the right structure to foul the functions non-coding RNA normally perform.

It's not throwing a bunch of car parts into a tornado and getting a fully functional car. It's throwing a bunch of random screws and metal shapes into an automated production line and having the production line grind to a halt or even start producing defective units.

For example, randomly cleave the poly A end off an mRNA, and over-methylate the guanosine (enzymatically not easy to do by accident, but chemically it is, such as when you're trying to form a nanolipid delivery capsule under conditions completely atypical for normal biological systems.) and you could have something that looks enough like snRNA that it could potentially mess with transcription or transcription regulation in unforeseen ways.