Hello! I missed this @-mention when it happened. I do have a experience building map tile services, and also keeping them updated (or, too often, failing to do so). It’s a fairly costly and complex thing to do. It’s not mysterious or cutting edge, but it is rather niche and specialized, with tooling and workflows that are pretty specific to mapping, map data, and maptiles.
One key issue for maps at iNaturalist is that we display a LOT of them. Generating, updating and serving maptiles at our scale would be a large effort with substantial costs, even if the data itself is free from OpenStreetMap. This older post, already linked in this thread, gives some details, which are still mostly correct. In one of my previous jobs (Stamen Design), we hosted freely available OSM-based map tiles at scale. From the beginning, we couldn’t afford to build an update pipeline to keep up as the world changes, so tiles would get increasingly out of date. Eventually, the company couldn’t afford to even serve the non-updated tiles at the scale they were being consumed (billions of tile requests), so they partnered with a commercial service to sell access to the tiles (and keep them current with OSM data).
The iNat platform also includes location features that are not visibly maps. For example, the place search on the Explore page includes a Google geocoding tool to interpret addresses/text into map locations. Over the years, I have tried many different options for that particular service, and Google’s has always done by far the best job of turning text into locations (which is surprisingly hard to do well).
We’re not currently considering any replacement of Google Maps or Location services.