Some possible incentives for running nodes (other than the ones named in your question):
People who want to run statistics on live tangle data will have a much easier job if they have the data locally in a full node (and get all changes pushed periodically) than when they try to query another full node (or have to wait until somebody provides a database of historical data to download and analyize)
Having adopted IOTA as a company, by running your own nodes you can easily avoid all the SLA issues and depending on others to run nodes for you. In case more than one company has adopted IOTA, it will still be decentralized
Those are not strictly economic ones (which you can do to "earn money"). But in case IOTA gets adopted, there are some possible economic incentives as well
Hosting nodes on behalf of companies that adopted IOTA and providing some SLA to the companies (getting paid by them) - similar to today's situation of companies who run OpenStreetMap Geocoding Servers¹
Running permanodes. Currently there is a permanode running (thetangle.org) which provides its data free of charge, but there are already plans of changing that. And with larger tangle sizes, and users who are interested in historic data, it is not unlikely that this happens.
Last but not least, if (and that is the bigger if) Qubic gets finalized and adopted, there will be direct econonomic benefits of running a Qubic node (assuming you can share the computational power required), which will probably perform better if you also run the IOTA node for it
In other words, as long as IOTA is still experimental technology and no company adopting it, other than technical interest (which you may subsume under "fanclub") or statistic purposes there are not many incentives of running nodes. And I can see that this is a vicious circle - as long as nobody has adopted it, nobody else will adopt it either.
That being said, I do not believe that the IOTA Foundation has a (direct) interest in trying to keep the network decentralized. They will do so if it helps to get their technology adopted, but if they can get it adopted without having it fully decentralized, they probably won't care about ensuring centralization and/or removing the Coordinator.
¹ Those are also known to be very resource intensive (for the whole world - i. e. planet.osm - expect about 1TB of hard disk space and about 1-2 weeks for a full database reindexing on typical datacenter hardware), but there are companies that are hosting them for companies that don't want or are unable to host them themselves and don't want to pay to other, more expensive, commercial geocoding services.
