Caliber 3035 - starting problem

[TheHitcher]

I’m servicing a very worn 3035 caliber and I’m having problems resolving an issue.
I’ll give you the issue first and then what other issues I had that are now resolved so you can dip in to find out what has been done.

It takes at least five turns of the ratchet wheel to start – the escape wheel tooth will stay on either impulse plane.
Most likely it is the gentle movement while winding what starts it rather than the force of the mainspring/train.
When wound and running, pulling the crown out and pushing it in, sometimes it won’t start.
The escape wheel tooth will stop on the impulse plane of either pallet stone.
I have seen it once stopped and the escape wheel was not touching either of the pallet stones.
If it stops on either locking plane where it should stop, it will start.
At times it picks up amplitude immediately, sometimes it takes around 20/30 seconds to pick up the amplitude when the crown is pushed in and it starts.
Seeing the escape wheel not touching the pallet stones when the crown was pushed in,
the issue is somewhere along the train.
Each train wheel was tested separately under the bridge, puff of air gets them spinning nicely.
Assembled train without the pallet fork in place, a gentle touch of the crown will get the train moving.
Tested without the cannon pinion, the problem is not on the dial side.

The initial issues:
Barrel scraping barrel bridge and ratchet wheel underside of the automatic bridge
Upper bronze bushing worn out – upper and lower bushings now replaced
Barrel and arbour worn out, too much side shake – new barrel complete in place and a new ratchet wheel
Lower chaton, the steel casing was cracked – now replaced.
Third wheel upper jewel two hairline cracks and hole out of round – replaced
Third wheel and second wheel upper pivots slightly rusty, new wheels in place now.
Minute pinion, plastic bushing a little worn out – too much sideshake. Now replaced with a new minute pinion with a jewel.
Escape wheel, too much endshake – adjusted
Pallet fork, locking depth unequal – new pallet for in place, both sides equal
All replacement parts are new and original Rolex.
Letting power down, the balance would keep going for 20 minutes with the click out of the way.
Almost resolved with replacing the sliding pinion and winding pinion – the balance now goes for about a minute. Stops much quicker if the crown is in second position.
Train bridge and pallet bridge are very difficult to put in place.
It has to be done with extreme care as they literally have to be pushed down.
It runs well. Running a long-term performance test over several hours, amplitude is steady. Nothing unusual showing in the log.
Long term test over several hours run in two positions, both fine.
Timing is within specification, fully wound and half wound.
It runs for 50 hours.
When it’s running, it’s absolutely fine.