So I have revised the PSU side of the board to feed 9V to the ESP which is probably better anyway to give the 3.3V more headroom to work with
Had an old ā66 that I drove to school from North Hollywood to West Covina back in the mid-late ā80s. Blew two engines.
Well since we have had the bus we have put a new engine (largly stock 1600DP) in and would like to look after it.
I could actually go the other way and use another switching module to supply the 3.3V directly? It has a higher current capacity than the onboard regulator which gets warm
The advantage of a stepped system is that the 3.3v will almost exceed that voltage, since it has an input rating of up to 12v. If there is sufficient heatsink (if the tab is connected to a copper filled zone of sufficient size) it should easily suffice, mind you also the ESP32 will make some use of the PCB as a heatsink.
I doubt you need to prevent back feeding 5v from the USB for the sensors, they won't draw much current, will they ? The USB can easily provide 500mA, with the ESP32 peaking at 280mA that leaves 220mA.
From my MegaSquirt experience,
Grounding becomes a critical issue in the car.
Insure there are no ground loops in the sensor ground path.
The entire Sensor ground path should be isolated and obtain its ground from the Uno.
In other words there should be no continuity to ground on the sensor ground path if the Uno is unplugged.
Shielded wires should be grounded at one end only to prevent ground loops.
Twisted pairs employed where applicable.
Nothing but sensors (5v devices) on sensor ground, do not connect 12v devices to sensor ground.
12v devices should have a separate ground to the same ground source as the ECU.
I think you will find the problem in this area.
After this I would, look at
try to avoid getting a coil negative trigger source anywhere near your CPU.
I have found a coil neg trigger to be one of the most difficult signals to isolate/cleanup.
Can you mount the Opto Isolator outside the case?
Can you instead Trigger off an ignition module tach output?
Given that the project is fluid and a combination of stripboarding / birdsnesting I have come up with a plan to provide something a kin to a ground plane. That is to add copper tape to the back of the board and solder to ground.
I have also found some ESD safe conductive ABS with which to 3D print the housing. It should offer a "little" extra protection.
Just waiting the watchdog chip to turn up so that I can do this outside of the ESP. That way if the ESP fully locks up the external watchdog will catch it.
Given that I'm just providing instrumentation its not the end of the world if I get a occasional reboot.
I have also had serious noise issues around using USB in a car.
I found that RS232 was almost flawless as an alternative.
The low signal voltage level of USB is a problem, in RS232 the signal voltage is much higher and less susceptible.
Cheers Rick, apart for powering the cab display USB is not used in this case. ESPNow is used to wirelessly transmit the data from the engine bay to the display in the cab.
I have built 3 sets of gauges and (maybe just lucky) but ESPNow has been flawless
Currently wiring an external HW watchdog to reboot the ESP if it locks up
Forgetting the fact that I have many of the skill sets missing here 
, what would a good power stratergy be for this?
My thoughts...
- incomming supply reverse connection & TVS protected.
- A step down convertor (9v) to Vehicle referenced electronics each with thier own LDO regulators (5V signal conditioniong opamps & ADCs, 5V to (Tiny Wideband module) and logic inputs
- An isolated step down convertor (9V) with LDO regulators, one deadicated to the ESP, one to the interface chips (logic side) and one to the signal side.
- All inputs to the ESP will be either opto isolated or bus isolated