regarding cross-coupled output stages:
they depend on the receiver treating the signal as if it had come from a transformer with a floating winding.
But the proliferation of gear having the simple chain-of-inverters line driver, that imitates a transformer with a grounded centertap, requires the user to exercise some degree of intelligence when interfacing with this circuit.
Grounding the low-side of the line is permissible if your ground system can tolerate having high-current distorted waveforms going thru it. Most can't.
So the prudent way to interface with this circuit is to float t.he unused side of the output.
But the floating-balanced output (i.e. cross-coupled) circuit depends on having the low-side tied to ground, and then feedback causes the gain to change, eliminating the 6dB change caused by not having 1/2 of the output signal. You still lose 6dB of headroom, since you're losing 1/2 of the output swing. Some of these circuits demand on having the low-side of the output grounded at the source, or instability results. I'm told that there are other variants of the circuit that don't care where they're grounded, doing a better job of emulating a floating balanced transformer winding.
But AFAIK, all of these circuits can't tolerate having the low-side float when talking to an unbalanced input... The usual result is signal loss, and in some cases wonky frequency response, but not like you'd have with a floating transformer winding with one side not grounded.
For my money, the best, most universal output circuit, is a floating transformer winding. There ought to be a law against unbalanced inputs.