I'm hoping that public nodes will run most or even all the current HAF apps.

As for public access to a HAF server for development purposes, as you suggest, this may also be feasible, but it is probably best to run it on a node that's not acting as a regular public Hive API node. We're basically planning to do that with one of our HAF servers (well, we've done it already once, planning to put up a new one soon) that uses the query_supervisor as a rate-limiter to protect against rogue queries. A multi-regional one might also be useful to reduce latency during app testing.