Re NMEA sentence variance: Perhaps you missed the part where I said I was doing statistical testing of receivers. For reasons I won't get into, the theoretical minimum jitter of the NMEA sentence is a bit under 1 ms. The VK-162 shows an ultimate accuracy of about 1.6 ms to 2 standard deviations. I hardly call this "all over the map" in talking about the time accuracy required for WSJT.
Re other serial receivers: Yes, I am aware of the Garmin units and several others, all of which cost more than the $50 device cost limit I used as a basis for my study. Is the $68-$15=$53 difference worth an accuracy of better than 1.6 ms? I would think not for Amateur operations. And as I have said before, if you have a Raspberry Pi you can get a PPS HAT for $29.95 which will keep the accuracy to the microseconds. Or pop $870 for a rubidium standard...
Re firmware upgrades: I think this is irrelevant as no one is purposing any new GNSS systems and if this does happen in the future it will have no effect on existing satellite information. The only thing currently being done is regional positional augmentation (SBAS) , such as the US WAAS and Japanese MSAS, none of which have any effect on timing.
What I expect to happen that will be interesting is USB 3 receivers that will have PPS capability. As I have said before, I have found one receiver that claims this but does not seem to be available in the US, no in depth documentation, and no idea of it's cost. And yes, USB 3 can have more than one data stream, unlike USB 2, so PPS is possible but at this point there are no device specifications and any talk about potential accuracy of a USB 3 receiver with PPS is pure conjecture.
Jim - WB6DKH