Based on the Google Streetview images, the northbound highway=pedestrian should be just a highway=footway and the highway=pedestrian is ending before the southern sidewalk along Zahntstraße.
Interesting. That is a very strict interpretation of where highway=pedestrian does and does not apply. Far be it from me to question how things are done in Germany, but given that the intersections I’m asking about are not in Germany perhaps we can broaden the discussion to regions of the world that do not have such specific signs.
Typically highway=* classifications change at a node in the middle of an intersection for simplicity even though technically the intersection is an area, not an exact point. A highway=residential intersecting with a highway=primary is generally not split so that the section within the intersection can be tagged highway=primary even though it could be argued that the traffic regulations of the primary road apply within the intersection, not those of the residential road. It would seem odd to me for highway=primary to break this pattern and require a different highway=* value when it passes through an intersection.
another way to look at it is that every road wide enough for vehicles in such zones are pedestrian highways, while those too narrow are footways. There’s no sign to say a single way is pedestrian, this is correct.
we do it for simplicity, but technically highway=pedestrian starts at a point if the road is represented as a way (or line, if you look at an area), and if it matters we should be as precise as required for a good representation.
Similarly, there are many highway=service and service=driveway with access=private which should be split for the part from the gate to the middle of the street so they are used for pedestrian routing on sidewalks
I would not call it strikt, that’s just a higher level of detail. I’m not sure, I would split the ways like mentioned above, simply because it’s too much effort. But that doesn’t mean someone else can put the effort in. Splitting the highway=pedestrian at the sign and continue the way as highway=footway. As I understood, that’s what the discussion is about. Someone put the effort and enhanced the level of detail.
So in many countries the edit in question is ok, but there are also countries where it’s not and the US might be one of the latter.
In my opinion the only difference between highway=footway and highway=pedestrian is that the pedestrian can have area=yes in OSM. If a pedestrian area is intersected by a highway type for vehicles, I see no problem in cutting it in two pedestrian areas and connecting them over the crossing area with a linear highway=footway|pedestrian.
If it’s really and explicitly one pedestrian area, and the crossing vehicles effectively drive as guests on a sizeable portion of the pedestrian area, I expect would expect explicit traffic signs, or comparable things like driveway kerbs, zebra markings over the whole crossing area, and yes, I would think that this crossing section of the pedestrian area would be a separate highway=pedestrian polygon with area=yes and e.g. vehicle=yes.
The crossing vehicle highway should probably intersect a both ends with the pedestrian area, and the crossing section could have a tag expressing that it’s pedestrian priority territory.
Yes, I agree, but this is not quite analogous. We split a highway=service + service=driveway so that we can tag one section with access=private and the other without. We don’t change one section to a different highway=* value. It remains highway=service to the center of the street. This is why I expect highway=pedestrian to also remain highway=pedestrian to the center line of any intersecting street. Splitting the portion within the intersection so that different attribute tags can be applied seems fine, but I don’t see why the main highway=* tag should change.
I agree that splitting and tagging the way going through the intersection with different attribute tags is a higher level of detail. The only part I find strict is the assertion that the main highway=* tag must change rather than using attributes.
in the cases around here, it is because the pedestrian highway will officially start at the border of the street as shown in the example above, it changes because a different legal situation applies. If details matter for the situation (or you like to add them) this is the representation that seems closer to the legal situation.
for pedestrian highways exceptions are somehow implied and it is expected that width is sufficient for delivery vehicles, while footways tend to be narrower and exceptions are an exception.
Does the law in your area say anything about driveways? Do they officially start at the border of the street, the middle, or somewhere else?
If the highway class comes with implications, why shouldn’t the highway class change at exactly the spot where a sign indicates those implications?
E.g. this motorway to trunk transition is added where the road markings for railway crossings are at.
Legally they would start at the edge of the road, though this edge is represented in OSM by the highway-way in the center of the road. So as long as you don’t start adding area:highway, the best representation would be to start the driveway at the road-way.
Obligatory: 
In the U.S., this was the subject of countless edit wars and mailing list debates until around 2021, when we settled on transitioning the highway classification at a nearby intersection if there is one.
One of the reasons we can’t take the signs too literally is that their precise placement is usually influenced by various pragmatic considerations such as sign spreading. Highway engineers are not allowed to post too many signs at a single location; they have to move the less urgent ones forward or backward some distance. At a street corner, the signs may need to be set back from the curb to accommodate turning vehicles or ensure a better line of sight.
The best example of this phenomenon is that a stop sign is usually posted past a crosswalk, but a driver or cyclist must stop before the crosswalk to let pedestrians cross. Similarly, traffic lights are usually hung past the intersection, rarely directly over the middle of the intersection anymore, because of rules about line of sight.
Even in California, the only state that explicitly signposts the start and end of every highway=motorway classification, we sometimes have to use common sense about the location the sign actually refers to. It can be difficult to maintain these nuances though. Despite valiant efforts, everyone still has their own ideas about the essence of a classification.
would you do the same with oneway=yes? (I hope that answer is no!)
we had service=driveway2 !!fun!! already… service=driveway is not for legal status but for functional one
Are you changing the maxspeed at a motorway only at the ramps? (I hope that answer is no!) 
But to answer your question: Yes, there are oneways that start somewhere between to intersections to allow access to some driveway and I would start the oneway at that location and no, I feel not obliged to map all the effected driveways.
not sure what the relevance is here, but when there is sign directly after road junction I set speed limit on road centerline since junction node
and Way: 1491778255 | OpenStreetMap joins road centerline
I meant sections like
that I see equivalent to ending highway=pedestrian at exact sign location
would you make it so it is without oneway=yes
But legally the maxspeed starts with the sign (and starting from the sign there is a grace distance and after the grace distance police can enforce the speed limit). If you challenge your speeding ticket that’s what counts. As I sad further above. It’s fine to simplify the reality for OSM. But you should respect, if others put in the effort and split it where the speed limit legally starts.
I see no value in doing this kind of splits just because sign stands on grass and would eliminate them.
I see value and did it myself if speed limit sign increase is not immediately after crossing but after say a bus stop
Again, you don’t need to see any value in that. I don’t see value in abandoned railway tracks or any wikidata tags or copying over ATP-data , still I should not delete them. Others sees value in mapping the maxspeed, where it actually starts.
