Coulomb

It looks like your battery is 16S LFP, so change your bulk charging voltage from 54V to 56V. You might even go higher, up to 57.6V, but overshoots could cause disconnection. 

You can connect to any phase.
Yes, but you need 3 "paralleling" kits and you set the three inverters to 3P1, 3P2, and 3P3 output modes. You can also add more inverters, with some paralleled on the one phase. For example, a fourth one set up as 3P2 and connected to L2, and has current sharing cables to the other inverter set to 3P2. Don't use current sharing cables with inverters on different phases.
The inverters have to have very similar if not identical firmware, and have to be models capable of paralleling (so none of the VM models, e.g. Kodak OG without any extra characters like plus). All Kings can be paralleled (and three-phased).

This is an Axpert King 5K-48.
The Kings are different to other models in that they are double conversion; power can be converted from AC (utility or generator) to DC (the "bus") to AC (for loads). So it can switch from battery to utility instantly, and can blend utility and solar power.
There is no reason that the battery can't be charging from solar when the utility is off. Of course, if you don't have enough solar, the difference has to come from the battery.
As posted elsewhere on this forum, the manual is here: https://www.mecerpc.co.za/pdf_downloads/10/102168_PDF2.pdf
The disc only has Watchpower, and very likely an old version of it. Some people think Watchpower is poor, others are happy with it. It can be downloaded from the Voltronic Power website, or slightly easier from here:
https://www.mppsolar.com/v3/download/ and choose your operating system under Monitoring Software.
No real limit. Above about 8 you need a special gizmo, I forget what it's called.
To parallel them you need one paralleling kit per inverter. These kits connect the inverters to each other via special cables; they talk to each other over these cables. Each inverter should have its own breaker between its AC-out and the load. It's not a trivial thing to set these up correctly, and there is a chance of blowing them up.
It's possible to split the loads and not run them in parallel, but then you have the hassle of running out of capacity on one inverter when the other is lightly loaded. Plus, it's two (or more) systems to configure etc.

Ah. The Infinis are not the same as Axperts; Infinis are grid tied, Axperts are off-grid.
The Infini-V models happen to have many of the same parts as Axperts, and happen to look much the same. But they operate differently.
Perhaps start a new thread with "Infini V 5K-48" in the title, to see if others with that model can help. I don't know much about the Infinis.

It looks to me like a hardware fault in one of the machines, e.g. a power supply collapsing, causing the DSP to reset after a few seconds.
Is it possible to isolate the inverter units from each other? It might be possible to run on one inverter (being careful with power) till you can get the matter fixed. Usually there will be two breakers connected to the two AC-outs. If so you could try to explain to your wife how to try disconnecting one, then if that doesn't work, disconnect the other.

That fuse is too light, unless you intend limiting your inverter to 3500 W maximum (70% of rated power). I'd use 100 A or 125 A. However, limiting the power will be kinder to your battery.
Those are a bit light. Your battery is rated at 200 Ah (at the 10 hour rate), so that's 0.1 C. That's cautious, but the battery seems to be an AGM type, which I believe can take a bit more current. I note that the datasheet states that the maximum charge current is 50 A. I'll leave it to lead acid experts to say what these maximum charge current settings should be set to.
I calculate 57.0 for 30°C as follows: 14.4 V per module or 57.6 total. But for 30°C take off 5 mV/°C x 5°C x 24 cells = 600 mV = 0.6 V; 57.6 - 0.6 = 57.0. I'm assuming that 30°C is a not-uncommon temperature in summer. You can increase the voltage in winter.
I calculate as follows: 13.65 V @ 25°C x 4 modules = 54.6 V. But for 30°C take off 3 mV/°C x 5°C x 24 cells = 360 mV = 0.36 V. So 54.6 - 0.36 = 54.24 V, closest allowable value would be 54.2 V [ edit: was 52.2 V ].
That would let you run your loads for the longest time, but would murder your batteries quickly. Unless you have something else to save the battery from such a large discharge, I'd put this at 48.0 V (possibly the highest that the inverter will allow). If you want a different balance between surviving load shedding and battery life, choose a value in between.

Oh. No, probably not, sorry. So it might be a long search...
For what it's worth, transformers are double insulated. There is the resin coating the wire (both primary and secondary), and the plastic bobbin separates the two windings.
As long as the winding wires are physically separated to they can't touch even plastic to plastic, it's pretty safe.

I've had suspicions that some paralleling cards can't take the extreme summer temperatures. Is it possible to squirt one of the paralleling cards with a "can of cold" when this happens, reset if necessary, and see if that delays the problem coming back? Repeat with the other one.
I forget the proper name for the "can of cold"; something that's electrically insulating and evaporates quickly, cooling the area.

A tin star for the Plonkster
Yes, the SCC and the DSP (Digital Signal Processor) both measure the battery voltage independently, and send their readings to each other. If the SCC finds that the readings differ by more than a certain amount, initially 0.2 V I believe, then the SCC will adjust its readings to agree better with the DSP.
 
Most of these values are stored as decimal numbers, with an implied decimal point between the last and second last digits. For example, your 00502 value represents 50.2 V, presumably the battery voltage as read by the SCC/MPPT/solar charger.
The SCC charge current I believe is no exception. Yet the current is always displayed as a whole number, i.e. they divide the number transmitted by 10 before displaying it on the LCD, thus throwing away one decimal place. Changing this is on my list of things to consider in future firmware upgrades. (Of course, the number may prove to be too noisy or inaccurate, but I say show it anyway.)