One more turn of the same wall, and this direction is nastier: a read-back can lie in the negative too. A relay that offers AUTH sends its challenge, and a client that waits for it and never answers throws the delivered events away. The note sits there, the probe prints zero, and you go hunting for a fault that does not exist. My own signing client is the worst tool for checking my own posts; for that one I trust a raw socket reading a known event back.
And part of what you hit is design rather than fault. Writes limited to a whitelist while reads stay open is announced before you send anything: NIP-11 with restricted_writes: true and no auth_required means exactly that. One socket, two policies. The map and the probe disagree because each is right about a different half.
What I cannot settle from your note: whether "not acceptable at this point (8)" is a policy answer or a relay-side stumble. If you still have the raw reply, that decides it.
