Saab 93 1.8 backfire only on startup

Hi everyone!
I don't usually ask for help on forums, but I have this car that has been kicking my butt for a while.

Saab 93 1.8, belt driven.

This is what happens. Upon startup, it will run fine for about 2s, and then the idle will bounce up and down AND, most importantly severly backfire into the intake... This will only last for about 3s and then you have to wait another 8h for it to act up again... so you can see that I have about 3s every day to diagnose and catch anything suspicious.

Generally speaking, I focused on the causes of the backfiring. Obviously no codes present...

I don't have any experience with backfiring, but I checked the usual.

I'll explain what I've noticed when the symptoms occur.

Fuel psi fine, doesn't bounce when backfiring.

Car V stable. Thought it might be a faulty alternator causing problems.

No fuel leaking from injectors. Even took them out and the spray pattern was fine. Left the rail out in the open and no fuel leaked.

No problems with the COP assembly (verified with scope current and secondary patterns). Even changed the whole assembly, changed nothing.

Cam/crank sensors fine when looking at scope.

Timing also fine. I even suspected a stretched timing belt (which doesn't happen as with a chain, but I'm out of ideas here).

Knock sensor was disconnected during startup. Changed nothing. Even disconnected it whilst the backfiring was happening. No change.

So far, every single thing that could relate to a backfire has been checked. Ignition, timing, knock, lean condition (smoke test), leaky fuel into the manifold, everything.

Even the ECU was replaced with a known good (pays off to have a parts changer boss sometimes.

Even the crank sensor was replaced.

So I suspected something mechanical...

The thing is, I can even crank the engine over with the spark/fuel disconnected for a good while, and it will STILL backfire when I put everything together. Thus, eliminating low fuel psi, valves not seating correctly on first startup, belt tensioner stretching over night, etc.

I've noticed that the throttle body changes it's position erratically during the backfiring, but swapped it with a used known good and nothing changed.

This is so frustrating, not in the least because I have 3s to catch anything.

Is there anything else I could check? I've run out of ideas! haha.

Thank you.

There's also no VVT, as I suspected perhaps a leaky solenoid that would change cam timing.

Truth be told the engine is as simple as it gets, yet it's really frustrating.

Based on this waveform, I changed the VR type crank sensor. 

Think that solves it?

Absolutely not.

The flywheel looked fine, but a flywheel problem would produce a similar effect all the time. 

P.S. what would cause the VR sensor to sometimes produce a small "swing" as in the attached photo? I know zooming in on a picoscope can make you overanalyze things, but still. It looks as if it was struggling to produce V sometimes. 

Unless those are the backfire events.

Ouch! Don't love SAAB, always a headache for me... I may not v
Be much help on this one, but didn't want to leave you hanging.
What year is this thing? I read the whole post thinking it was a 1993, lol.
I think you covered most of the bases, popping through the intake is usually lean or timing. How's fuel quality?
Hard to see the scope set up, but changes in amplitude and frequency on a VR type sensor usually are the result of changing speed of the trigger wheel. Are these captures during the back fire?

Ah, Noah, thank you for stopping by, This car is driving me insane, especially that it's a NA, no VVT, as simple as it gets engine.

It's from the early 2000s, I'd I'm not mistaken. I dislike working on Saabs as well!

Yes, the scope capture isn't the best, and Im actually certain those changes are caused by the backfiring... Which is why changing the sensor made no affect.

So timing or lean? Timing was verified, it's spot on and belt driven, so no slacking like with a hydraulic tensioner.

Fuel quality? I think it's fine, good idea, but it drives fine.

That leaves me with a lean condition... But I checked with a machine. Ill try spraying some carb cleaner and watching trims, but I don't expect to notice anything wrong.

Timing cannot change after having run good for 2s... Fuel isn't pooling in the intake, no lean (so no fuel being taken away or no vacuum leak), nothing in the cylinders after the night, I can crank the thing for a while without coils firing to verify fuel is getting to the injectors (if by any chance it wasn't, which it is) and also unstick a sticky valve if there was one...

Unless a valve might stick/open when the car has been running but that's really, really unlikely. I unfortunately don't have an in cylinder transducer.

Sticking a vacuum guage into the manifold won't help.

Perhaps a MAF? A bit desperate, unplugging it makes the car surge, so I'd need to source another for testing. I think the values hovered around 2g/s which is what I would suspect to be for a small engine like this.

At one point I forgot to bolt down the fuel rail, and I saw puffs of smoke getting past the injectors every time it backfired. That was an all cylinders, so it's something that affects everything... Like fuel, timing, air... Not one sticky valve...

I'm at a complete and utter loss here haha!

That crank signal looks crazy. Make sure that the trigger/reluctor wheel is not loose or floppy. 
Add an ignition event on a second channel and sync it with the crank signal. Does the ignition timing act erratically at start up?