Motors stop but TinyG thinks everything's OK

As I was reading the top of your post, I was thinking noise issue. So as I reached the bottom of your post, I see you are already suspicious of grounding issue as well.

From what I read, you have turned the entire machine structure into an antenna for noise that will couple into the limit switch inputs.

Here are suggestions for limit switch wiring

- Use shielded twisted pair
- Connect the shield at the tinyG end to the power supply negative input on the tinyG. Better to make that connection directly at the Power In "-" rather than on the switch input terminal switches; noise currents flow directly to the Power supply (an input filtering) rather than across the tinyG board
- Make no connections at the limit switch end except the switch terminals, do not connect the "gnd" wire of the twisted pair to the machine structure itself.
- You may find that using shielded cable for connecting the 4 stepper wires
for each motor will help. Ground that shield at tinyG end, to Power supply negative as well.
- Shielded cable for your spindle might help as well. If problems persist, using a separate power supply for the spindle might be appropriate

I'll add that bonding all the machine metal together an then to ground is probably unnecessary. but if you choose to do it, connecting to the power supply earth (green wire in line cord) is probably best, or to the DC negative at the power supply output, but not both.

Transformerless AC to DC power supplies typically do not have a direct connection between the + and - outputs to that earth AC input.

I thought a bit more on your problem and wondered as well, you did say that motion terminated but did not say it was a flashing leds/ limit exceeded sort of event.

My guess would be that the coupled noise somehow affected code execution in the tinyG controller, although explaining how the program execution could remain 'sane' is a bit of a stretch. Best guess would be corruption of the dynamic part of the process (RAM values)while the tinyG FW and parameters, in non-volatile storage remained in-tact.

I understand your baffle.
In a typical machine, none of the active devices (steppers, limit switches, spindles, etc) should have any electrical connection to the machine framework.

If you wanted to pursue understanding what was going on, I would start by attempting to understand in detail how your power supplies are connected.
The Low cost (relatively speaking) compact transformer-less DC supplies typically used in this space have a indeterminate relationship, output wise, to AC Line-in and Neutral. Some have their positive output sitting one one two diodes above at below Line in, while - output 24V below Neutral. Way back at your breaker box, Neutral and earth (green wire) are connected.
But other designs are just the opposite, - out is referenced to Neutral and + output 24V above neutral.
Very strange things can happen if you connect the wrong things together.

Thanks for the follow-up, some good suggestions.
This is a world where hobbyist techniques, connectors and layouts meet industrial applications.

I'll add to your list a caution that dealing with the typically fine gauge wires that connect stepper motors to the rest of the world can be difficult to work with and maintain. A forum suggestion a few years back suggested using wire ferrules, I found them very handy. Here is what I used:
Wire Ferrules

I suggest not connecting either Power - or Power + to Earth ground unless you really understand ALL the elements in your system.
There are no universal standards for implementation of transformerless power supplies. Look at and read the schematics of your subsystems.
Using optically coupled switches for interfaces like the tinyG output port that controls spindle motors can help. If your system uses external stepper drivers that connect to the internal logic ports on the tinyG controller, you will find that virtually all the available off the shelf units are opto-isolator coupled