All the more reason not to expect mappers or data consumers to rely on width measurements to express whether something overlaps something else.
From my perspective, this is almost correct, but a parking lane can still be implied by the provision of ample space. In the U.S., signs are only present in neighborhoods where curbside parking can be restricted in any way; markings are only present where a poorly parked car would be a collision hazard due to limited horizontal space; and the only reason for a surface change would be a concrete pad to prevent heavy delivery trucks from damaging the asphalt pavement. Otherwise, they’ll just pave some more asphalt and call it a day, yet the result is functionally equivalent to implementing all this additional infrastructure. I don’t think we can rule this out in other countries either.
To reiterate, if two can validly drive side by side, it isn’t a single lane. If one can validly drive beside a parked car, it isn’t a single lane. This is not about physically blocking a lane, which depends on so many specific factors, but rather about the social construct of mandatorily maintaining a lane as a road user – two orthogonal concepts. On a conventional street, this only affects cyclists and pedestrians by analogy to motorists (or else we would count bike lanes in lanes=*).
I think this is a good practical concept to maintain in OSM tagging, versus a rigid global standard that relies too heavily on rules from one part of the world. We need an obvious way to record this social construct. Otherwise, the earlier joke about lanes=2.2 might not be so silly after all.
If we want to additionally indicate whether parked vehicles impede moving vehicles, we need a series of subkeys, not just a few values in a single key. For one thing, motorcycle parking isn’t necessarily going to impede car traffic or vice versa. There’s also a temporal dimension, most notably local rules allowing delivery vehicles to block travel lanes. And bike lanes are often blocked by other obstacles such as garbage bins on Tuesdays or long-term homeless encampments. This is all interesting and useful information, but the use cases are mostly distinct from the use cases for lane counting.