What do you mean when you say "lured into an attack chain"? If the false chain is does not have more chainweight than the true chain, then you would have to prevent the checkpoint server from accessing the open internet, a feat that becomes very difficult if we implement multisig. If it does have more chainweight, then it is by definition the true chain, so broadcasting it is not an issue.
I'm referring here to the "long-range attack" or "history attack" - e.g. the attacker would buy a large stake of coins, deposit them on his wallet (so he can cryptographically prove he "possessed" them), and then sell them again, but secretly minting a chain where he transfers the coins he supposedly sold to an address he controls via a double spend.
This chain would need to be "designed" so that it accumulates more chain weight than the "true" chain (so it's not a cheap attack -> he, in some moment, would have to possess a very large amount of coins, but he can sell them again).
If he can manage to spot the IP address of the "checkpoint server", I suppose he can try to lure him into a reorganization using his attack chain on multiple "fake nodes" that would try to connect to the checkpoint server.
It would be very probably worse than a normal history attack (where the attacker is trying to "lure" large minters into his chain) if most nodes still follow the centrally broadcasted checkpoints.