Thanks for the helpful info.
I've removed the 6A diode from inline with the supply and placed it in parallel with the 3A diode across the motor. They are no longer getting hot.
Somehow this has stopped the resetting problem also???? Any ideas why?
Beaniebot's post #14 gets to the heart of it. But I'll add my two farthings worth. I'm in a type-ative mood.
Your power supply diode was the likely culprit for the worst of the resets. Any spike voltage higher than the power supply voltage was isolated from the low impedance of the power supply output and was free to create what havoc it could with the rest of the circuit. Removing the diode in the power supply lead allows the supply to sink as well as source voltage/current, thus stabilizing the load and helping suppress noise spikes. It's necessary to stabilize BOTH sides of the circuit's power supply.
The paralleling of the snubber diodes merely helps spread the inductive energy dissipation chores - not a bad thing per se - but not the fix for the reset problem. BTW, if one diode is slightly faster than the other - almost certainly - it will pass more current, making it warmer than the other, dropping it's junction drop voltage below the other diode's and stealing more of the load - creating an imbalance problem. Which is why you don't see paralleled diodes in well designed circuits.
Using a Shottky diode for the snubbing job is a common practice and highly recommended, especially if you don't have access to a 'scope to see what the noise spikes look like. Shottky's have a lower forward V drop and faster reverse voltage recovery - they're just plain better at handling the inductive spike. When you can't see what the possible problems are it's best to use the components that are most likely to solve or avoid problems.
I'm in agreement with the suggestion to put a substantially larger capacitor on the input to the 5V regulator - not to provide more current to the regulator - but to absorb voltage spikes from elsewhere. I'd also suggest adding a low value cap on the input to the 5V regulator to catch high speed spikes. The larger a capacitor the worse it is at catching high frequency spikes. You might also consider adding a small resistor in front of the regulator input capacitors to further suppress any spikes that might be feeding the cap. One that dropped a tenth of a volt or so in normal operation should work well in conjunction with a large enough cap to spread the RC time constant over several pulse cycles.
That would effectively decouple the regulator from any likely input spike.
Without access to a 'scope - in a non-production build such as this - I'd put a fairly large value and a very small value cap at both the input and output of the regulator. (And anywhere else I thought it might help.) When you're working blind, overkill is good.
Finally, (as noted by others,) careful attention to an effective ground system with no loops is essential. Use a separate heavy ground on each high current load and take each lead all the way back to a common tie point at the power supply. Star grounds are your friend.
Time to stop my rant. I've already said more than I know.