@vepkenez and his rowdy band of miscreant node spinners have observed that the first node in a deployed fleet has trouble discern the fleet state perceived from other nodes.

This is because we don't absorb fleet states from endpoint requests (it might be worth discussing whether it makes sense to start - it might help #2152).

However, in the medium-term (perhaps as part of the domain fixes in #1881), we do need to adjust the keep_learning_about_nodes behavior so that the first node in a fleet starts asking for fleet states as soon as its public_information endopint is hit (with well-formed POST data) for the first time.