The second most common flaw is harmonic distortion. In this case something in the detector system through audio stages generates harmonics of the detected tones. This causes extra bars to appear spaced at multiples of the detected tone frequency. It is often blamed on the transmitting station, but if the extra signals are multiples of the tone frequency it is a receiver system flaw.
I have never understood these linearity issues (which is really what they are) designed into radios at audio frequency after most of the gain has been applied, The hard bit is the IF stages not building a clean audio section!
If you sold a radio with 1% harmonic distortion in the IF amp (or worse the first mixer....) at normal operating level, people would rightly be screaming that it was a steaming pile of merde, but apparently audio frequency problems are acceptable?
The third most common issue is intermodulation, which requires mixing of multiple tones or signals.
However, unlike the others this one can occur in the first mixer, so the strong signals do NOT need to be inside the passband.
so naturally the weak link is the transmitter and receiver because they are analog.
Not sure I agree with the conclusion there, for all that I agree that using an SSB radio for a mode that is not 2.4Khz wide is stupid!
transmitters would be directly digital modulated and receivers would have filters to match the TX bandwidth and do digital detection. It would not go digital to audio, audio to SSB RF, SSB RF to audio, and then audio back to digital.
Nothing wrong with using an audio frequency first TX IF or last RX IF (for that is really what it is), in fact this can have a few upsides as it removes a tricky DC nulling problem in the IQ modulator in the transmitter.
Also you really do not want to just apply raw digital switching to the phase modulator as you will get switching sidebands for days, much better to go to analogue (either at baseband I/Q or as a low first IF) and shape the transitions with a raised cosine to minimize transmitted bandwidth.
AD have an interesting combined DDS and I/Q mixer that takes its input directly in the digital domain that might be fun to play with for this stuff.
There is nothing much wrong with using analogue radios and even an audio frequency IF, providing the selectivity is there, but there is no point in trying to design a mode much narrower then the selectivity filters in the radio either, so something 500Hz wide makes sense, maybe even 250Hz but only a vanishingly small number of rigs have first IF filters narrower then that.