I am curious about something…what does the signal look like if the Jetson communicates with itself on that UART port in loopback mode? Connect TX to RX, and if using flow control, then connect CTS to RTS. Do you see delays in TX->RX like in the image you posted above? If not, then you might describe your cabling. I noticed a lot of noise in the yellow TX (which would be the other end’s RX).
That cable is unshielded, so quite possibly this would contribute to noise (not so much of a worry at 9600, but a problem at 115200, and definitely an issue at 921600). What is the source of noise in the yellow TX signal of post #9?
Incidentally, do the Jetson and other unit have a good common ground? At the moment of applying a test instrument which has a ground the test instrument itself might introduce a ground loop. Is this a drone whereby attaching the serial console might change the ground connection, or even attaching the oscilloscope might change grounding? There’s lots of information on the internet on ground loops, but I like the schematic illustration in this (the schematic notes the difference between a common with one ground versus several grounds which are otherwise not common): https://www.youtube.com/watch?v=gBCOeYfRL5g
Notice that your UART on the Jetson has to share a ground to some other system except when loopback has the UART talk to itself; however, adding the oscilloscope itself adds another ground. Do you see that noise in loopback on the Jetson? If not, then it is safe to conclude that the TX/RX/Ground on the Jetson is ok by itself. Similarly if the UART of the remote device is without noise in loopback mode. If the noise occurs only when the two units connect together, then you might have a ground loop (or at least a low quality common, or both).
This does not say that this is not a software issue, but until you figure out the noise I wouldn’t bet on a pure software solution. I’ll suggest shielded cable and considering if you have a ground loop issue.