I had space for decoupling caps on the PCB for each IC, but omitted them on this particular prototype. I had space on the PCB for 7 decoupling caps, the six 595 shift registers and one 74125 logic shift IC. I don't remember the last time I had seen latch up like this so I was curious how many decoupling caps or smoothing caps it would take to get rid of the issue. One or two decoupling caps and it was difficult to get latchup, but with all 7 decoupling caps installed (and no smoothing cap) I left the inputs floating for days and it did not latch up. The smoothing capacitor will likely make it even more resilient.
Some more info if you are curious. I had prototypes with both TPIC6C596N and TPIC6C596N (high power versions of the '595) and both would latch up. It varied how many of the ICs would actually latch up, anywhere from 1 to all 6. Each of of them would use up 360ma at 5v, or about 1.8w. I verified latchup was between VCC and GND. I left one latched for a few minutes and surprisingly it was not destroyed, I'm sure it's not good for it though! It needed a complete power down to reset.
Yes, keep an eye on your power supply noise. I know my lab power supply was not happy with the disturbances as it would blink the current overload (CC) icon, even though the slowly updating display showed it no where near the set current limit.
Indeed I was looking at what a nice thermal camera you can get now for a few hundred. I remember not too long ago we were amazed at the "inexpensive" handheld flirs that were coming out for *only* a few thousand dollars!
Regards,
-Moses