Personally, I feel the real test for any proposal is how well it works for mappers and various types of data consumers in practice. The stairs=yes tag has some decent usage numbers already, but there aren’t enough different classes of applications using indoor data yet to tell whether it will work without issues. I’d like to see at least 3d rendering and proper indoor routing implemented somewhere to validate the concept. (OSM has some difficulties with fixing mistakes once they have been “canonized”, in my experience.)