My 2c worth (forgive me if I explain what may be obvious):
It looks like your boards have isolation, perhaps motor controllers, and some high current / high-voltage stuff going on. I'd have a look to make sure something in this environment isn't killing your micro :
1) That micro has no clamps for positive over-voltage on the I/O. Thus it can be broken easily by any kind of inductive spikes or ground bounce. When a pin dies, try measuring resistance to ground and to +3V3. That may give you some clues. Also check the resistance values while the micro is powered up and while off. If you see a resistance path to GND in the order of kΩ, suspect a HV spike with low energy behind it. If you see ohms, something with current behind it has trashed the pin. See if you can still measure the negative clamping diode for the pin to check if it's still there. If you have inputs coming from off-board or high-current/voltage sources, isolate with a few hundred to a few kiloohms in series (Adjust R53 according to your needs), AND add a clamp to supply, and then another resistor in series as well (R3 can be as low as 100 ohms, but do your calculations assuming an ESD hit of 8kV from a 2k source) - if there is e.g. a negative going spike, you don't want the current shared between your protection diode and the internal protection diode. I use the scheme below for all inputs that travel any distance or to a pushbutton or LED anything that might pick up static from a user/machine. A BAV99 can take 2 or 4A for a microsecond and is almost free and has practically no leakage. If you can't clamp to your 3V3 rail due to too little current draw (micropower designs e.g.) then clamp to a shared transzorb, even if you have to pre-bias it to supply.
2) Check your grounds. Make sure there is no ground shift happening due to high currents in the groundplane/fill. Make sure both VSS pins are connected to each other very solidly, and to same GND as your input signals. If you have high current, make sure your micro is on its own, quiet, ground with no current flowing in it. And a nice ground fill underneath it if you can. If you're on a 2-layer PCB, you may struggle. Check for placement of switching regulators and other sources close to the micro, especially keep the SW pin away from the micro.
3) The input pins can't be damaged by a 3.3V signal injected when the micro is off, you'd have to go > 4V above Vdd to do that. But it can be damaged by negative going signals in excess of 20mA.
4) Suspect the firmware. If you can reproduce the fault, try and get debug output of the GPIO registers of the affected ports.
5) Check for EMC. If you don't have equipment for this, take a seek coil out of an old hard drive and wire to a BNC into your oscilloscope. Wave it around in the vicinity of the PCB and play "warmer warmer colder colder" till you can see what's coming out of where.
6) Check that reset line - I think there's supposed to be a 1k resistor in series with the cap on the reset line. And a push-button can be a nice place for static to enter a micro, so consider whether it needs a clamp.
7) Check the layout carefully, look for routing of all pins going to the micro. If they run close to high dv/dt nodes, especially for long runs, you need to shift them elsewhere.
Thank you for your input so far guys. Ill much appreciate any new ideas. Especially after a few weeks of being stuck...
Ive uploaded the PCB and a measurement of the Vdd supply.
