‘The scientific society pays less’ not if the number of papers explodes, which is the case. ‘Society’ pays for 10 incremental/redundant OA papers at 1500$ a piece instead of 1 in depth study. This only to help academics to get tenure in universities that stopped bothering about quality and just ask for volume, or well-paid government jobs.
‘Note that publishing a paper in non-OA costs well over 5000$’
This is a wild exaggeration, it has been calculated by the Harvard library that it’s on the order of 3000$, about the cost of 2 pay-for-publish ‘open access’ papers.
‘all publishers must switch to OA model’
I don’t see why they ‘must’, but most of them already have because of the high demand of authors that are desperate to publish their stuff.