Also, potentially, sending an update to browser might take too long and the TouchAction might will be assumed as 'auto'. This might cause issues if event handlers (whether preventing or not) take too long and the frame has had an effective touch action which was not 'auto', say 'pan-x pan-y'.

As Rob suggested one idea could be to potentially send touch action update twice; once a tentative one and then a restrictive one in case one of the handlers prevent defaults touch.

Also, if fixing 'touchmove' using 'touch-action' is problematic, we might eventually just force the 'touchmove' handlers as passive.