Our understanding is mongodb should scale to tens of millions of rows better than a SQL database (which would be a nice problem to have)

Once we made the decision to use mongodb, it made sense to use the built in near functionality, rather than roll it ourselves.

Indeed. I am using Django to build my app and a mysql db with 2 float fields for lat/long. GeoDjango and GIS extensions for mysql seemed like over kill, plus I don't really know what I am doing but it seems to work. Django doesn't have very good support for mongo. Well, the django orm.

To be honest I can see absolutely no reason why you can't use floats as you describe and < , >. Depending on what you are doing it may be a little tricky due to lat/longs being angular values, but over small distances I think this can be largely ignored. (In geosay we have some front end code that sorts out the niceties of the angular stuff)