One thing different about my 4048 which is may be why I've not had problems starting what motors and such I have is that I have it temporarily connected with about 800 Farads of Boost Caps (64V value of bank) connected to the bus bar from which it is fed. The capacitors alone could supply a multiple of the current that the inverter could possibly handle in the short run, it is in reality limited by the wiring connecting them to about 8000 Watts if the bank was at least 50 volts or so before, and about 7200 Watts (at 50 Volts) via a 150 Amp breaker for longer time frames. The capacitors though can only supply about 150 Watt hours between 56 volts and 44 volts when paralleled with a battery, although they can supply that energy in a really short time frame. When hooked to just capacitors, without batteries, it will run from 64V down to 32V which would supply about 340 Watt Hours because of that larger voltage window which would not damage the capacitors. You do have to equalize the voltages through resistance before connecting the battery bank to the same bus bar. I suspect the max power output from the inverter at 32V is probably greatly reduced even though allowed in the inverter setup. I would have a bit of concern about damaging the inverter at heavy loadings at that low of voltage level as Trace generally stated operating voltages from 44 to 64 volts if I recall correctly, although, that said, I have never had a problem with loads up to about 1500 Watts at lower voltages and have never ran anything more than that at those levels. Although I have run pretty fully loaded, aside from starting motors on appliances such as laundy and fridges, the only bigger motor I have run would be my air compressor which also does not mind the inverter, even when switched to just the capacitors and no batteries, although the longer term power would be limited to the current solar production under those circumstances so if your were going to operate it in this configuration for long it would have to be sunny out. My welder is only a 120V. repair welder, so it is not really comperable and I've never tried to run the 220V plasma cutter off of the inverter through a transformer. I have never tested, nor do I intend to, its ultimate surge capabilities as I need the thing to last a long time before I have to pay for repair or replacement. I think of the capacitors as more of a big DC buffer than an energy storage solution because they are limited to supplying an amount electrical energy for about one load of laundry using a gas heated water and a gas dryer in their storage at night if no batteries are connected. For now they are more experimental toys, although I am curious as to how long they will last.
At any rate, in Dan's case I think he has enough energy in the batteries to run the welder consistently, but he may not have the ability to deliver enough power to the inverters when they start. Although I read what was posted, I don't fully understand why one of the inverters kicked with two stacked and did not with one run through a transformer unless the transformer on the output of the single inverter reduced the peak loading at start up because of its inductance. As was pointed out, the more the DC voltage drops the more DC amps are required to maintain the power output, I think the example was 400 Amps at 35VDC for a 14KW surge, and capacitance across the DC inputs should reduce that short term voltage drop, thereby reducing the amps required for the surge. If the voltage held to 40V minimum, 14Kw would then only require 350 Amps on the DC side. For an instant at startup the inverter wouldn't act differently whether the power came from a capacitor or a larger battery bank, and the amplitude of the dip would be reduced in either case.
Just an idea, relatively easy to experiment with, and not likely to do any harm if it does not help, but probably only a few thousand additional microfarads of capacitance would be required as 2.5 Joules delivered in the first millisecond would reduce the peak loading required to deliver the equivelant power on the DC side by about 50 Amps or so. 6000 microfarads properly wired up across each inverter could do this with a starting range of 50V and an ending low point of over 40V after delivering that power for a millisecond. A 1000 microfarad 100V capacitor is a $1.59 U.S. item at Jameco, and is frequently found iin old power supplies and such. I have a few single electrolytic capacitors of suitable voltage range of larger capacity in with my electonic parts I've collected over the years, one was from a DC forklift. Six of the 1000 microfarad ones, paralleled though sizable enough conductors, if a single larger unit was not at hand would likely be relatively easy to rig and check. One would have to build a suitable structure to get the amps from each of the capacitors or capacitor banks to the terminals of each inverter though for that instantaneous surge. I am not sure if the current overload Dan's unit is showing is on the DC or AC side of the inverter though, and if on the AC side this should have no impact. At any rate, just an idea.
I agree with Chris that preferable in the long rung would be the load balancing transformer, as well as a bank and wiring capable of delivering the power that can be demanded by the inverter, but a lot of folks don't actually have the resources to do it right to start and a band aid is the only immediate option to making it work. In spite of not having the resources many still have the wife looking at them when the lights go out or stuff does not work, with that outstreched finger pointing in their direction, and are hence under a lot of pressure to get the thing working as well as they can with options available to them. Chris's wife is apparently lucky, as he had both the know how and the resources to get it done, others use bandaids, or suffer the ongoing wrath of their significant other.