Identification of possibly problematic OSM data

[501Ghost]

How can mappers help to smoothen out the OSM data so that changing over the data model to something new will not introduce a lot of errors, misinterpretations, lost data etc?

I have a few suggestions myself, although I don't know how relevant they might be:

• Fixing broken relations (e.g. relations without a type=* tag)
• Fixing overlapping nodes
• Fixing / separating ways of which the tags can imply both a linestring and an area/polygon (e.g. highway=* + landuse=*)
• Fixing problematic characters in keys

And how do we go about fixing them?
Tools like Osmose and the validators in editors are already available and we can come up with specific queries in Overpass, but how do we coordinate the task of improving the data specifically to support an update of the data model?

(I originally posted this as an issue, but it fits here much better.)