I've done it and seen it work, but you guys got me to wondering exactly what I might be compromising and by how much, so I went ahead and did some half-assed tests. I can't imagine it will convince anybody of anything, and I wasn't particularly surprised, but it was interesting.
So, I don't have a particularly deep mic locker, but I picked four dynamic mics that I use quite a bit, and I think are in the range of things somebody who can't afford a mic cable might be using - an EV Cobalt 6, an EV 635a, a Nady DM80, and (since a 57 was mentioned above, but I don't have one here) an SM58.
The best I could do for a consistent signal source was to play a song from YouTube through BlueTooth from my phone to a Gemini powered speaker. So not exactly a full range signal, but I think good enough given the limits of these types of mics. Completely not anywhere near scientific, but...
I recorded from the beginning of the song three times through each mic into my Tascam US-1641. Once was into the 2.2K XLR mic input with the gain all the way down which is -6db, and where I have tended to keep them just for safety's sake. Second was with the gain knob up to where the signals were peaking around -8dbfs. I did not try too hard to match these levels very closely between the different mics because I was just looking for an idea of how all the way down compares to more normal usage. Third was with an XLR > TS cable directly into one of the 10K Line Inputs. Those don't have gain, so just straight unity.
Then I normalized them all and ran them through SPAN for some investigatating. The difference in all cases is noise, but the result is counter-intuitive. The noise on the pass with the gain up is always significantly quieter than the other two, like pretty much consistent with the amount of gain that was applied by the preamp. This must be the noise floor of the interface, and must come mostly from after the preamp circuit. Without any gain, it's still around 60db down or better, which is better than cassette or any electric guitar, and could be better with a bit louder source.
It turns out that the TS input is always hotter than the XLR input at 0, but almost never by the full 6db that the specs would suggest. It is, however, quieter in terms of noise by a bit more than 6db in all cases.
But I literally see no difference in the frequency response until we get up to the >12K region where they all roll off pretty severely and the lower gain inputs are lost in noise.
So it basically all depends on everything of course. Different mics and interfaces will interact differently, and the only way to tell exactly how the gear you have will work together is to plug it in and find out. I don't ever understand why people go looking for more complicated and/or expensive solutions rather than trying the thing they have right there already.
Whether the noise is too much depends on the volume of the source and kind of the needs of the destination. It has convinced me that I should probably twiddle the knobs a little more when I'm recording with dynamic mics rather than just taking them at 0, though honestly with the things I record it has literally never been an issue.
I have never actually meant to come across as advocating the XLR>TS thing as the right way to do it. All I said was "with loud sources and sensitive microphones it can usually work", and that's pretty much still where I'm at.
