The fault is immediately on start-up
The wires are routed differently, I plugged them in in reverse order, but still have the polarity correct, I've double checked that. The return wire is actually shorter than when it was on the roof previously.
Same thing happens if the 110w is removed.
I was a little concerned that that might be the case
I will send a picture tomorrow
So either one string of panels hooked up is fine. Both hooked up causes the problem.
Return wire longer or shorter isn't what I was thinking of, but how big a loop it makes. 20' of wire forming a circle is a big antenna, large "loop" area. Same wire out 10' and back, maybe even twisted, makes a poor antenna, no "loop" area.
I'm not sure how two loops makes a good antenna when neither did by itself. This isn't a "rabbit ear" where each goes to one terminal. each loop connects to both terminals.
There was a time when SMA Sunny Boy had a lot of problems with AFCI, from comments I read on their forum. Yours worked fine before, presumably antenna formed by panels and wires is different. Maybe some kind of damage; a short to frame as Picasso suggests would show up as a fault during startup (if frame grounded). But each works alone. Each having a fault at a different voltage tap of the string, then current would flow between them and arc. Are your mounting frames grounded?
Confirm that arc fault is the error, not some other message given.
Transformerless inverters have no galvanic isolation between grid and panels. They do a test on start up. If used in a 3-phase (208V) system, the panels will carry an AC voltage, and capacitance affects them. Bad grounding of frame might even cause arc noise in that case.
"Sunny Boy's recommendation in the past has always been to disconnect and reconnect all the connectors which I have done several times"
So, not the first time, Huh? But I guess it has been behaving itself for a while?
Maybe it's that d*mn 5G? New transmitters in the area?