This is the fourth PM6-600 I know of that throws overcurrent errors specifically on A4, regardless of any motor and cable swaps, and if I remember correctly, most forum posts asking about this problem refer to A4 as well. Makes you think, ain't it?
When testing, remember that:
- KRC1-era robots control motor brakes in two groups of A1-3 and A4-6, can't open them individually and need to energize all motors in a group to be able to move just one, as the motors have to hold the axes in place when brakes disengage
- encoders have to be swapped when doing motor cable swaps, this is best done at the RDC (robot base, where the thinner cable from KRC goes in), otherwise the robot will go bonkers for a split second, then e-stop due to excessive position error
- it is safest to swap A4 and A6, both have no mechanical limits to exceed by accident
- A1-A3 are influenced by gravity and/or the counterbalance as soon as the brakes are released, but A4-A6 aren't, so you can leave a motor disconnected during jog testing of the other two (technically speaking, they are influenced, but the force is so low it can't backdrive the harmonic drives...)
In any case, it looks like you already ruled out any cabling, motor or drivetrain problems by swapping stuff around and/or leaving the A4 output completely disconnected and still getting the error. (BTW, no, you won't damage anything by leaving outputs unconnected)
On mine, which did the exact same thing and resisted the exact same attempts at fixing it, I swapped the power PCB along with the IGBT packs - I got two untested PM6-600 units very cheap on an auction site and cannibalized one of them. This helped, ruling out any problems in the logic board (on top of the power board) or the IGBT driver cards (in the middle). Before I found the donor units, I started probing stuff, both off- and online. What I managed to figure out is that the current sensing transformers on A4 looked fine, so it had to be one of the auxiliary components in the sensing circuit. This might be different for you. Then again, with this being yet another A4 overcurrent, I suspect a design fault in the power PCB causing a consistent, repeatable problem.
Now, remember, poking around in a powered, half-disassembled PM6-600 while it's connected to the robot isn't the safest thing to do even with personal protection equipment. Don't even think about trying it without *proper* PPE. You can and most likely will die if you do this. There's no galvanic isolation from the mains input in a PM6-600. That huge capacictor bank in the middle is charged up to ~550V DC (thankfully, it does discharge very quickly when power is cut). It will kill you a hundred times over without even noticing any extra load and keep running in front of your slightly smoldering, spasmodically twitching corpse. At minimum, buy certified insulated gloves. They're surprisingly cheap for how much safety they offer. SECURA ELSEC gloves should be readily available in France, for example. Use multimeter and scope probes with insulation running up to the very tip, to avoid shorting stuff. Get a grinding face shield (the ones designed to stop shrapnel from an exploding angle grinder disk, not the flimsy COVID things) and a welding apron, for when you *do* short stuff and get a splatter of the molten probe in your face. Figure out how to power and ground your oscilloscope to safely measure mains circuits, so you don't fry it. You can bench-test devices like this if you know what you're doing (guys at KUKA had to do it somehow), but it's no joke. Be safe.
And before you go any deeper, just look on eBay or whatever local French auction sites there are for cheap PM6-600s as replacements or part donors. They do show up.