(while I’m quite approving of your personal and inspiring story, here is why I think the second part of your post has many unaddressed problems)
Some would argue that this ATYL policy is the primary (if not only) reason why OSM succeeded, while countless others have failed. If one has to wait for some bureaucracy to approve a tag for use, it is bound to create horrible friction (and probably destroy much of the community in process).
Have you ever seen how it works in practice? e.g. look at RfC: Highway=Scramble - even after a months of discussions, no consensus have been reached. This is not isolated example by any means, just one of the more recent examples. And it is unlikely to be solved in this year either. How long do you think should be expected to wait until the thing that they want to mapped is “approved” and thus become possible to use (in such “prescribed-tags only” scenario), before they give up and move elsewhere?
I think a major quality criterion for a map is reliability, and for this consistency of mapping is crucial.
This I would agree with. (that is why I am vehement opponent of proposals attempting changing wiki definition once a tag is in widespread use; unless extremely well laid plain with strong understanding of enormity of the task is laid out first and followed methodically)
It does not however imply that the only possible solution is to have some authority approve what tags may be used, and what does may not be used. I would argue that such a “cure” would in fact be much worse than the original “poison”.
I think much more emphasis should be given to consistency of mapping, so every mapper should feel it is their duty to follow the consensus as documented in the wiki.
That would IMHO only be possible if wiki is forbidden to ever be changed once it was created (i.e. if it stopped being a wiki). Otherwise, you’d always have situations where some mappers use one definition, and some another (as they never heard of updated definition) - it is (again IMHO) impossible education problem to solve (finding all mappers and re-educating them, as well all old versions of tools using previous definition, all online and offline tutorials, videos, etc, and re-writing them to avoid new users being taught obsolete facts etc.)
The wiki should not describe mapping practices, but prescribe them
That is eternal “prescription or description” disagreement. It seems to be no easy consensus about it in any other part of human activities, so I see little hope consensus would be reached about it here.
Status quo in OSM seems to be descriptive approach - that way (at least) things can be documented, and data consumers can chose to avoid using tags that are too vague/ambiguous/problematic.
The onus is thus on proponent of the new method to convince others why status quo should be changed (and be forewarned that it is likely no easy task at all!)
after consensus is reached about it
And here we come to the crux of the problem, reaching consensus. I don’t have hard numbers, but it seems to me that number of wiki editors (even if combined with number of active tagging ML and forum users) is orders of magnitude less than number of map editors. So, even if that small minority were to agree on something, that would not represent the consensus of mapper (which are the only thing that matters, one might argue). And very often, even that small minority disagrees. Quote heavily, in many cases.
How would that be handled? Depending of all mappers to follow tagging proposals/suggestions? Ain’t gonna happen judging by the history, and assuming “well its their problem then” would just be enforcing current state when anyone can (and does! as most wiki editors will inform you) rewrite the wiki according to their will without any consultation with other wiki editors (much less this community forum, or, gods forbid, the majority of mappers - we don’t even have a way to attempt to do the latter!). So doing that would not help with decreasing (much less getting rid of!) that ambiguity that troubles you (and all of us).
Sure, there might be technical measures to try to enforce it and solve “low participation problem” - system might be set up which would for example disable editing capabilities of all mappers who did not read and vote on all tag proposals. That would be even more horrible and lead to massive loss of users who lack the time and will to be involved in such infinite time-sink with low reward-ratio. (especially take into account that with disabling ATYL policy, the number of proposals would rise exponentially)
If there is no consensus on how to map something, this means that some mappers will consider the activities of other mappers as “wrong”, leading to wrong data (in their eyes). I think we all agree that wrong data is worse than no data
Not really. If there are e.g. two ways how tag is being used, it creates ambiguity, and not “make data gwrong”. Depending on the specifics of that ambiguity, that tag might still have some (reduced) use (e.g. sport=hockey). Only in cases when the ambiguity is extreme (much rarer), then that tag can only be considered “burned” (and documented as deprecated for such reason).
Another more common thing that happens in ATYL (i.e. without someone “prescribed allowed tags”) it that there are multiple tags to describe the same (or very similar) thing. That is much less of a problem (just a minor one-time annoyance for data consumers to set up aliases).
Also note that even if only well-documented tags were allowed to be used, it would not solve the problem. Many are are least somewhat subjective (like the smoothness tag you mention, and have dedicated a lot of - appreciated! - time on), and so people can still mistag things (i.e. tag in a way that mismatches current wiki definition). The fact that meanings changed across the time practically guarantees that - some people will still use older definitions (as mappers don’t usually go read a wiki every day - they will usually read it just the first time until they understand it, and afterwards use if “from head” without re-referencing it again)
so we should stop mapping until we have consensus about how to map, and then all of us should follow that consensus
So, basically, you suggest that we should make OSM read-only immediately, and leave it in that state for indefinite future (I’m trying hard to avoid saying “forever” here)? 
TL;DR – It is extremely hard problem. Yes, I (obviously) agree that everybody agreeing on exact meaning on tags is best outcome. I however disagree that either having “central committee” prescribe the meanings of tags, or allowing few hundreds of wiki editor to mold a world in their image is a good idea. (even if they could all agree, which I find doubtful).
IMHO, the best we could hope to do is to is to document how things are used, and suggest better ways if current usage is problematic, and try to convince users to use more complicated and tiresome proposal process so unclear definition could be caught early enough so they can be fixed, before they become problems. But I’d dare not outright forbid ATYL, as it would lead to loss of much of community, I’m afraid, which would be much worse than the original problem (ambiguity of some tags).
in one of his posts on the StreetComplete Github. I’ve been much more active posting there since I started using SC when corona came. My doctor recommended me to walk a lot every day, and SC gives me the destinations to walk to.