SteveD

It seems to me that both Pylontech and Voltronics are to blame.  Pylontech for requesting voltages that are too high (in the real world, where 2nd rate manufacturers implement poor control algorithms that overshoot) and Voltronics for being such a 2nd rate inverter manufacturer.
 
You will probably find the other batteries are all in advanced stages of degradation.  Consider setting battery type to USE and max charge voltage to the lowest that still gets you to 100% - my Pylons work well at 52.6V.
Also consider setting the float voltage down to about 50.5V.  It will mean that your SOC will drop to about 99.5% as soon as the batteries are full, but it will greatly enhance their lifespan.
Of course using an Axpert means that your inverter cannot be relied on to correctly switch to float, unless you are running a version of @Coulomb's patched firmware.

The latest firmware for your model is over a year old now, but it has the "presumed good" MPPT logic (so the "stuck at 90 V issue is largely solved).
It's 46.82, available here: https://forums.aeva.asn.au/viewtopic.php?p=102981#p102981
Edit: You may want to update the display firmware as well, as 112.19 is quite old. The latest in the y12.xx display firmwares is display firmware version 12.28.

I say you cannot set the generator value. It sees it as grid and from grid it can at any time charge what your setting 11 is. The inverter doesn't know you are using a WEAK grid. 
You will have to manage the grid charge to say 15A for in case the kettle is on. 
My own view and I could be wrong. 

I don't believe that your model can do what you want (reduce charging power to achieve a power limit at the AC-in port). It's possible that the Duplex models (with two AC input ports) might do this; I'm not familiar with these.
You could maybe do it with clever monitoring software: it notices the overload, sends a command to the inverter to cut back or stop utility charging), but it may not be fast enough. By the time that the overload is detected, a decision made, the command is sent, and the charge current is reduced, the generator may have been overloaded for several seconds, and may well have tripped by then.
The inverter could probably limit the overload to a fraction of a second (say 50 milliseconds), which may be fast enough. 
[  Edit: Duo → Duplex ]