That makes sense, it forms a simple snubber network. A capacitor in series with a low-value resistor might work even better. Did you try a freewheeling diode directly across the valve leads?

Yeah, I think you’re right.

Only thing I can think of, maybe it’s a bleeder resistor for that cap, and it failed by some kind of internal short which reduced its resistance (and increased its heat dissipation hence the blackened board)? But fails-short is an unusual failure mode for a resistor and 1 GΩ is pretty high even for a bleeder, so maybe we’re misreading something.

That’s interesting, so you can flip the relays all you like without trouble as long as the 24DC supply isn’t connected? If that’s true then your problem presumably isn’t the typical inductive kick from the relay coil. It looks like your relay board has stuff on it which is presumably drivers and snubbers so let’s assume all of that is adequate to the job.

So, if it’s inductive kick from the valve solenoid it’s being coupled all the way from there, back through the 24DC supply to the outlet, then forward through the USB supply to your shift register, which is impressive! But not implausible.

Anyway, three places I’d add some stuff:

• The main thing you need is a snubber network across the valve solenoid coil itself, ideally physically close to the valve (you want to minimize the area of the loop formed by the valve coil - wiring - snubber). Something as simple as a freewheel/clamping diode would probably help a lot. This will also improve the lifespan of your relay contacts which are probably arcing a little.
• Small decoupling cap on your breadboard, say 0.1µF on the power supply rails, to keep your logic happy.
• Larger decoupling cap on your 24VDC rails (the bus on the left), just to eat any transients the snubber doesn’t deal with. Maybe 1-10µF or so?

It mostly doesn’t matter.

If it’s a high-current, high-frequency, or low-noise circuit then maybe the inductance or resistance of those traces would matter, but they’re very short so probably not.

If you’re mass-producing it, then sometimes the reflow or wave solder process works better if the traces leave the pads in particular ways. You’d talk to your manufacturer about this.

If this is a hobby project, you’re overthinking it; arrange them in a way that pleases you!

From that description it sounds a little bit like the CUI PJ-096 ? Not a common connector type AFAIK

(Found it via connectorbook.com)

In addition to the voltages being different between real-RS232 and “TTL”-serial, they’re also swapped. On a DB9 you probably have something approximating RS232, where mark=-9V and space=+9V, but the debug header is likely mark=+3V and space=0V. So even if your inputs can handle a wide voltage range, the sense is inverted, which is why you’ll get garble.

(For example, when the line is idle it’s at the ‘mark’ voltage and the receiver knows a character is incoming when it transitions to ‘space’ for one period (the start-bit). If mark and space are swapped, the receiver will see ‘space’ most of the time and only detect a character starting when there are some ‘mark’ bits in the middle of a transmitted character. It’ll never actually synchronize correctly with the transmitter.)

You can figure out what you’ve got with a multimeter and checking what the voltage is on the TX pin when it’s idle.

They might be willing to spec it as “quiescent current” (current drawn at 0 load) even if they don’t provide a full curve. Annoying that it’s not on the datasheet.

I’ve had middling-to-good results making battery contacts out of springy bronze metal stock. It solders well, it’s easy to shape, and if you get the right kind of metal it retains its springiness well. (510 or 544 alloy, maybe? It’s been a while.)