I made some experiments where the nozzle is swapped upside down so that gravity cannot pull out a string.
My theory is, that with retracting, some air gets inside the hotend which is caught by surrounding, sealing melt and warmed up then, pressurizing the melt again. Therefore my strategy would be a fast, short retract of about 1mm to remove printing pressure, followed by a very slow retract (for about 15-30sec with about 0.5mm/min and step-wise descending speed) after swapping to the experimental idle position, which should compensate expansion of trapped air or outgassing moisture/other fumes.
I already did some tests with a macro and it reduced the length of the emerging string significantly (about 3mm instead of >10mm) but still not 100% and therefore, I will do some further testing in the next few days.