Upgrading 2 planes to 3.7 causes the Castle ESC to triple beep "ESC detected unusual signals or loss of signal on throttle wire" unless I boot in failsafe mode

Very weird, one plane had a pixfalcon with 3.6 (I’m pretty sure it was 3.6 and not 3.5) and once I upgraded it to 3.7, every init with a radio link gives me this Castle error on the Talon ESC
"• • -" “Radio glitch” "ESC detected unusual signals or loss of signal on throttle wire."
on a castle Talon.

Then I upgraded my other plane that was running 3.5 to 3.7 too, and now I’m getting the same init problem with a different Castle ESC (Edge 50).

The way around it is to init without a radio link. I turn off the LRS receiver, the plane inits in circle/failsafe mode and whatever it outputs on the throttle channel (using PWM, ESC plugged directly into the throttle port of the pixhawk).

So it seems that AP 3.7 outputs a signal on channel 3 throttle) when the radio link is working, that upsets the ESC at boot/init time and apparently that wasn’t the case with AP 3.6 and definitely not 3.5.

Both planes are using openlrsng receivers. Is there a a chance that maybe AP 3.7 passes through data from the receiver more quickly than it used to, and it’s when it’s still in some low value failsafe mode, and that in turn ticks off the castle ESC because the value it too low or something?
Well, not quite, I went into the openlrsng failsafe menu and changed the throttle failsafe from 804 to 1200 and that did not help. If the RX is on when I turn on the pixhawk and ESC, the ESC goes into this error mode.

Now the only way I can boot my plane is to turn off the RX, boot the pixhawk/ESC with the RX off, let the ESC init, and then turn the RX on. The plane then gets out of AP failsafe and the output of the THRO channel goes to the ESC and everything works fine.
Except that I didn’t have to go through this whole dance before AP 3.7

Looks like it is a regression in AP 3.7 which is fixed in AP 3.8.
I’ve opened this issue in the meantime: