osm2pgsql-dev
Optimizing geometry storage
Storage-Optimizing PostGIS Geometries describes a neat trick.
It should be really simple to build this into osm2pgsql. Question is: Is it worth it for our data?
I'd be interested to hear about your test results using ST_QuantizeCoordinates...
My preliminary testing using ST_QuantizeCoordinates even resulted in a dataset of 14GB ending up at 15GB if I can trust DBeaver's relation size display (woodland data of OpenStreetMap), which is the opposite of what I excepted. The data was CLUSTERed at the end of processing, so I am sure there is no table bloat. A few other datasets gained only a marginal reduction in storage size (maybe 2-3%). Admittedly, my workflow has some more gears going on, but even taking that into account, I have a hard time explaining such result.
Some thorough testing seems warranted though, as it appears your mileage may vary depending on table contents.
