Aaah. I found your other photo. [I expected it to be in the same 'album' as the first]
I suspect that the feed wires to the input of the A-R unit are a weak electrical service connection. Wiring to the input side of the A-R unit should be directly connected to the booster, NOT the constant polarity track leading up to the reversing loop. The difference??? A matter of the total electrical resistance in the current path, and therefore short circuit current available for the A-R unit to quickly sense a polarity mis-match and make a correction before the booster trips out. Using alligator clips with typical small gauge wire for testing doesn't help.
The A-R unit will naturally/always start up with the output phasing in the same direction. [Satisfying point B as you describe it.] Once the poor electrical connections to the A-R unit or to the track cause a delay in tripping a polarity mis-match at point A, causing the booster to trip, It will never recover as when the booster trips, operating power to the A-R unit is lost, causing it to restart when the booster recovers in the normal startup position... with matched polarity again at point B and the mis-match again at point A. Flip the polarity to the input of A-R unit to test my theory to see if now the issues occurs at point B.
Making your wiring more robust should certainly help... including good wire feeders near both sides of the isolating gaps on each rail.