No disagreement that it's a bad long-term idea (as I've already stated). But absolutes like "NEVER" are hard to defend. Emergencies do arise, and for those, it's SOMETIMES okay.
How can this work? First, the media properties are not as different as you seem to think. The coercivities are within 10-20% of each other (e.g., 650 vs 720 Oe), and the active-layer thicknesses are not significantly different, either. These differences, in fact, are about the same as the variation from vendor to vendor, and over temperature, to say nothing of the variation in write fields produced by a random assortment of drives. So, it can work ok (notice that I don't assert that it WILL work ok; I try to choose my language carefully). The recommendation I mentioned -- that you use only virgin floppies (and proper degaussing is a good "revirgination" process) -- greatly enhances the chances of success, allowing you to get up and running in a pinch. If all you need is to get an OS onto a hard drive, this can be a lifesaver (I've done this several times, so I'm not speaking theoretically). If you're expecting the floppy to retain that data reliably beyond that time, you've got high hopes.
Going the other way -- using an 800K as a 1.4MB disk by cutting/drilling/melting a second hole -- similarly CAN work, but is inadvisable for anything but a short-term emergency hack. And again, should it work once by some miracle, be happy with that luck, and don't expect anything more.