I agree to all points. For point 6. Mapbox is a big contender for ESRI's current position.
For 8. Openstreetmap, GDAL and associated open source ecosystem offers a solution to proprietary format and API of the month problem to a degree. It is not surprising that there is no money flowing if mapping is not a core business of the users.
Aerial and satellite imagery is the heavily regulated and lucrative part of any mapping solution. If you are allowed and able to obtain that, you can probably get by just selling the raw imagery without geospatial processing.
Mapbox tries to replace Google maps, ESRI is different category which is only slowly challenged by enterprise vendors (Azure and other clouds) and FOSS: PostGIS and other sptially enabled databases which have more and more geo features, making specialized GiS to a niche.
I write and maintain some OpenLayers API (OSM) Javascript, used in a product.
OpenLayers was selected after Google started charging 10x for their maps API, grayed out the existing tiles, then required a credit card on file. Since the API key can be "view sourced", that was totally a non-starter, along with the always-changing ToS. Way too many moving goal-posts.
(Part of the reason non-free is a problem is that any testing ends up costing money, and the more testing you do, the more $$, which is a disincentive to do testing. But just staying on top of lopsided ToS gets old fast when you can just replace it.)
Using a combination of OpenLayers and SO sample code, I've been able to do everything I wanted to do and am happy with the final result, but usually it's some of the toughest programming of the month.
For 8. Openstreetmap, GDAL and associated open source ecosystem offers a solution to proprietary format and API of the month problem to a degree. It is not surprising that there is no money flowing if mapping is not a core business of the users.
Aerial and satellite imagery is the heavily regulated and lucrative part of any mapping solution. If you are allowed and able to obtain that, you can probably get by just selling the raw imagery without geospatial processing.