As for audibility, we have to consider the RMS noise (20Hz...20kHz) when looking at hum/buzz spectra, and other artifacts as normal distortion, idle tones etc.
An example, this is the Apple USB-C Dongle (EU version) in two setups differing only in the balancing current flowing through 0.5m of 3.5mm to 3.55mm cable before it enters the measuring ADC. One time it is zero, the other time the injected mains leakage was that of an Intel NUC PC with its class-II SMPS supply flowing through the cable into the class-I Notebook.
The RMS noise floor was measured at -108dBr ref 0dBFS at the input, again displayed as dBV, sorry). A single sine at that level would have the same energy, so it is certainly audible when embedded in that noise, assumed we listen loud enough to actually hear the the tone and the noise floor. So we have
some chances that it's audible occasionally.
The shaded area is the level below which a single tone within that noise fades being audible even at highest playback levels downstream, this happens at around -20dB. This is the more or less guaranteed
ignore area for any noise and distortion even under most stringent circumstances (maximum gain after the DAC).
In this case most of the dirt is below that and so can be safely discarded as definitely inaudible.
Given that the actual level of that noise usually is at or even below the hearing threshold most of the time (otherwise normal playback levels would deafen you in a split-second), the chances of audibility are even less, in this specific example. Equal Loudness Contour Curves at work here, too.
But in a more complex setup hum/buzz etc levels may quickly reach audible levels, though. Typical is a PE-grounded PC with a soundcard feeding another class-I device like an amp or mixing desk etc. With separate mains outlets and several meters of RCA cable this more often than not an unusable connection, IME.