That would imply that once the '1' is established, the strong pull-up is not reactivated if the pin is pulled low externally.
correct
If you look in "the bible" you will see that the "strong pull-up (P1)" is operated from the port latch. Whatever external sources do to the pin does not affect the port latch.

For purposes of input to a port there is one simple way to do it" "always make port pin input happen from a push-pull CMOS source" e.g. HC/HCT, memory, FPGA ...

I would revise the design so that the speed sensor was connected to the input of a schmitt trigger (e.g. HC14 [two in a row?]) because
a) the HC(T) chips can take a lot more abuse (+-20mA) than a uC pin.
b) you will have the noise resistance of a schmitt where the uC has none
c) a RC filter to a schmitt input is stable, to an uC input it is questionable
d) the characteristics of a HC(T) input are well specified (1uA max), the characteristics for input use of a quasi-bidirectional port are, at best, fuzzy. One example: what happens if your input is held low by some undercharacterized device and the pulsed strong pullup blows the "undercharacterized device".

just a footnote which may and may not apply to your chip:
I got 'caught' by a derivative that did not have the "pulsed strong pullup" and had to add external pullups to get a reasonable response for a port that had to be quasi-bidirectional (the excuse for no starng pullup was that the port in question could be configured as push-pull)

Erik