Elsewhere it was suggested to me to use area:railway=tram instead of landuse=railway, which does play nicer with surface=*. Doesn't solve the sedum problem though, since none of the established surface values fit that either.

First off, I'm open to any ideas on how to mark the sedum track beds better. I've searched in other cities that use sedum on tram tracks, but they all either tag it as grass (which it absolutely isn't), or not at all, so I just went with Spr3's idea. One of my other ideas was landcover=flowerbed + species=sedum, or something along those lines, but I'd like more opinions on this.

Adding landuse=railway is a separate thing, while not strictly necessary, it's technically correct on tracks which don't double as something else (road, ped way). I am, however, aware that this isn't done often (personally only found it in Krakow) and if more of you feel like it's for some reason bad, I'll undo this. If it's alright though, I'm ready to extend this on all exclusive tram tracks in Bratislava to keep the city consistent.

My previous work resulted in institutes with multiple close-by buildings as multipolygons, which I at first decided upon after reading the wiki page for relation:site, which lists alternatives to said relation. To me it seemed like a straightforward way to mark that two buildings both house an institute.

I asked other mappers and was advised to either tag both buildings with the same tags, which I wanted to avoid to make maintenance of the institute itself easier.

Further advice was to do what I did with this changeset, which is to set an area around what I know to belong to each institute and tag it with its information, leaving the buildings as they are. I went with this, because it also lets me leave the addresses on the buildings themselves. I'm open to any suggestions to improve the model.

Note: I decided against reverting, because when making the multipolygons, I also refined the building parts as well. Simply switching from MP to buildings and yanking the MP's tags that belong to the institute to the new areas was the simplest way without having to do all the work again.

France includes many overseas territories, so when you do a bounding box for France, it's pretty much half of the world.

The new facade is mostly guessed, but I have asked its makers whether it would be possible to release any kind of plan to the building, to make it more accurate. If that fails, I'll try to make do with a rangefinder to improve it.

Oh, got it, it's that overlap. Okay, that makes it a mistake on my part, should have noticed it's crossing a part of the building that's not there (as in not on that level) physically.

You're right, there's an entrance for the restaurant, which technically is an entrance into the building, but not the office part. StreetComplete doesn't make this clear, I'll take more care next time when dealing with entrances, thanks!

An honest mistake on my part, thanks for spotting it.

Interesting, I didn't edit these tags directly, but via StreetComplete, so I guess this should be reported as a bug in their repo, if it produced incorrect tags.

Right, thanks for the help, I'll remember to check register adries as well next time.

They're not proper buildings, but rather a shipping container and a container house with fast food joints inside them. In addition to no number displayed (which in itself doesn't mean anything), they don't show up on cadastral maps, therefore they don't exist as buildings before the law. So they can't have numbers. There is a small chance that the cadastre has erroneous data, but I went with the assumption that they don't.

is this line of thought wrong?