I am destroying microcontroller ... dual power to VCC: switchmode and USB)

[Odd]

If, while connected to USB (which feeds VCC via a diode)  - I connect VBATT ... then VCC will overshoot to >8 volts!

What is the solution to this issue? a diode from USB to VCC and then another diode from switchmode to VCC ?
I can not do USB+ via diode to VBATT  - because the switchmode is very noisy (ripple) then Vin is less than 0.8v above Vout.

I could rework everything to run of 3.3v ?

[Gribo]

Your GND net might have a problem, The one that goes into the MCU is not explicitly connected to the one with the ground symbol.

[wraper]

Connect USB connector GND/shell and all of MCU GND directly to the ground plane and do not bother doing such ground separating nonsense ever again.

[wraper]

What is the solution to this issue? a diode from USB to VCC and then another diode from switchmode to VCC ?

Dunno if this has anything to do with the issue but most of switch mode and linear regulators do not like backfeeding. As of when it's outputting voltage I guess it could cause instability in its feedback loop.

[Odd]

Some regulators do not like beackfeeding and get hot (this one does not)
To me (and my scope) it seems like the switchmode regulator kind of lives and it's soft-start is perhaps down.  It may even be that it is at nearly 100% duty cycle.
Then the VBATT skyrockets, and the regulator overshoots badly.

the groundplane is not an issue, GND is properly connected and there is a copper fill (GND) layer.

Right now I think about having a diode between the switch mode and VCC  (and increasing it's Vout a bit to compensate for the diode.)

[Gribo]

There might be current from the switch mode regulator going through the MCU and into VBUS.

[Siwastaja]

Also inrush due to cable inductance ringing with the 10uF very low ESR ceramic (C3), see https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

[ajb]

Does it overshoot when connecting VBatt without USB power connected?

Just because the regulator doesn't get hot when being backfed does not mean it's okay with being backfed! There are many possible routes for backfeeding to cause a regulator like this to malfunction.  The bootstrap circuitry, soft start circuitry, internal error amp, etc could all exhibit problems when starting up while being backfed.

Adding a diode between the output of the regulator and Vcc would be a good idea.  There may be other solutions, like changing the way you handle the enable pin, but the diode will almost certainly solve the problem as long as the converter otherwise works. 

[Tnixsw]

Maybe the USB power is upsetting the feedback path for the ME3116.

A diode after C11 to isolate the switch mode power from USB power?

[Odd]

thanks to all.
I've solved it right after my last post by redesigning and adding another diode after the switchmode, (and increasing it's voltage to compensate for that) - so both diode's cathodes feed the VCC.
- then I forgot to tell..

Before solving it I troubleshoot it with a scope. it became very clear that if the VCC was fed without the step down controller doing it - it's soft-start feature was disabled, or - it's PWM got really big (as it probably wanted to regulate to the set 5v, could see only a lower voltage - and increased the pulsewidth to ... a lot)   - then - when VBATT was connected .. the first pulse was much too long.