This seems to be the most comprehensive explanation of the cause of false positives from the loopback test on some CH340-based boards:
Additional discussions:
- Starting here: Arduino failing loopback test despite working perfectly otherwise - #5 by kprims
- Starting here: Sent lines different from GRBL echoes · Issue #845 · grbl/grbl · GitHub
- Starting here: Problem with Nano clone from China - #10 by walker1
I have personally reproduced this phenomenon on my CH340-based boards.