Debugging joint amplifier fault

My machine (Kinematics XYYZ with 4 servos) has been running without a problem for months, but today I started getting "joint 2 amplifier fault". It happens quite quickly after starting any job now. Joint 2 is one of my dual Y axis joints.
I have not found any docs about what this error means. Any hint on where I should start debugging? I have not made any changes on the hardware nor software.
 

Turns out my joint 1 servo (not the one that is named in the error) is not working anymore, leading to joint 2 having trouble moving the gantry.

Checked the cabling of the non-working servo, all good. That leaves the servo (JMC with integrated driver) or the interface card (MESA 7176e) as a potential source of the problem.

It gets more puzzling:

• If a swap the two Y axis motors, they both work - if I swap them back, only one works. Repeatedly tried with the same result. In any case, both clearly have power, because if I try to move the non-working one manually, it tries to move back 
  
  . I am confused. Could this be servo tuning related?
• In addition, on my destroyed workpiece I see that there was a large following error on the X axis, not just on Y.

I don't see a common denominator. Except for the MESA card, but it seems to me it would be a strange failure mode, too.

Did you swap the motor connections at the 7I76E or the motor?

If differential signalling is used and you have either:

A bad cable
A bad 7I76E step output
A bad motor driver step input

You can get flaky behavior if just one on the differential
signals is working.

I swapped them at the motor side. I also made sure to check every wire with a beeper while moving the cable around, all wires seem fine. But I only checked the non-moving motor wires, not the other ones... maybe I'll check them all.

I just checked both channels with an oscilloscope:

• Once with the servos disconnected.
• Once with the servos connected, such that one doesn't work.

Each time, I measured both channels directly on the servo connector.

When disconnected, channel 1 looks like this:
 

When connected, the pulses are way lower:
 

As a control, I also measured on channel 2 (when connected):
 

I also measured directly on the MESA connector in disconnected state, but both look the same in that case.
So I guess this means that there is a problem with the driver of the JMC...

A lower signal is expected especially with a differential inputs that typically
have 100 Ohm termination.

Did you check both  STEP+ and STEP- (assuming you have differential connections)
 

Lower yes, but why the large difference between both servos?

The above measurements are between step+ and step-.

So the large swing difference is traceable to just one drive? (and not one step/dir output)?

Good question. The bad motor on the good channel is also pulled down a lot, but only to 3.7V instead of 2.4V.

The specs of the servo say they tolerate a signal voltage between 5V and 24V, so both is well below that.

I am really not sure who is at fault here; the MESA because it cannot provide enough power, or the driver because it pulls too much power.