Motors don't like high frequency inverters. While it's a bit of a shot in the dark, it sounds like everything is at this point, so consider adding an autotransformer to the system. The additional inductance may provide the amount of filtering needed to get the inverter's algorithms in line. I'm not sure this will do much, below is my napkin analysis, but I've found that a lot of off-grid systems, particularly high frequency, don't have enough inductance in their system to make every load happy, and an autotransformer solves a surprising amount of weird, annoying, but non-critical problems such as the one you're experiencing.
Going back to fundamentals, the "inverter" is actually an AC/DC (probably reversible) inverter, and a solar charge controller. If the inverter isn't reversible, then an AC charge controller may be found as a 3rd component as well, but this is rare these days. While there's only one interface, the reality is that these two components are controlled by separate microcontrollers, with separate firmware, different operating modes, different operating frequencies, different priorities and alorithms, etc.
Given that your problem only seems to occur when the power is coming primarily from the solar, then the problem is likely the interaction between the inverter and the solar charge controller.
If you think of them as separate devices, then you recognize that at the power level, the only interaction they have with each other is on the 48VDC bus.
I'm guessing that if you put an oscilloscope on the 48VDC bus you'll find wild spikes at the inverter terminals, and these spikes are causing the inverter and/or the solar charger to switch between operating modes thousands of times a second as the apparent battery voltage changes. This, in turn, results in audible noise as the capacitors vibrate while they absorb and discharge the spikes as much as they are able.
If that's the case, then the solution is to stabilize the DC bus.
Unfortunately, with an all-in-one, the DC bus between the two interacting components of the inverter are inseparable, and can't easily be brought out and filtered. The capacitors in the inverter are intended to provide this filter/buffer, however in cheaper inverters they typically don't use high quality capacitors, so they are slow to absorb and release energy, allowing particularly high frequency noise through.
So let's focus on what we can change. Tell us about your DC side - what are the lengths and gauges of the wire going between the inverter and the batteries? What are the capacity and rated input/output (Amps, continuous) of the batteries? What, if any, components lie between the actual cells and the inverter (include BMS, fuses, disconnects, wire/bus transitions, etc)?
The solution may be increased battery capacity (you have at least 200AH, right?), higher quality batteries, decreasing the resistance between the cells and the inverter (thicker wire, fewer connections, star arrangement rather than daisy chain, anti-ox coatings, tighten the connections), or perhaps a high frequency, high power capacitor setup at the inverter battery connection. Do NOT put ferrite or other filtering anywhere on the DC lines - this will make the problem (if this is the problem) worse.
Or it could be none of these, but having read what you've tried and how your system is setup, this is the route I'd pursue were I in your shoes.
