I guess it depends what you’re chaining. On our second TP bot, our lift was chained to the motors because it allowed us to locate them in a convenient spot (don’t ask why we had 2 lift motors). In that case we very much needed proper tensioning in order to accurately control the lift’s position. Our original cortex bot had a drive/lift transmission, which meant we needed either a long geartrain or chain. Tensioning was less of an issue there, as we had encoders on our wheels, but it was still very helpful in order to prevent wind-up on the velocity controller.
Of course you should never make chain too tight- that’s why we tension in the first place. Our lift, for instance, would have worked with one less link, but it was extremely tight and added way too much friction. We added a chain link, but this introduced a ton of slop, so we tensioned it in order to be as tight was possible while not noticeably increasing friction.
I guess you would consider this incorrect building, and I agree that it’s usually better to find solutions without chain, but in many instances it simply makes things much more practical and, if you can eliminate the primary issues with it (slop). I would say it is perfectly acceptable to control things with chain. Our lift worked extremely well. We had no issues with slop due to the tensioning, and no issues with friction in spite of the tensioning.
In your example of linking mechanisms that are already driven directly (I know many teams who did this for their chassis), you’re right that slop definitely isn’t a big issue. I would have thought this is an unusual use case though; in fact, off the top of my head, I can’t think of anything but a chassis in which you would do this. (I’m sure there are other use cases, I just can’t think of any right now.)