The error logs certainly help, cheers for that. Think I might have an idea whats happening here and its certainly not to do with the tokens. Just from the logs
D 17-08-05 23:12:33.589292 [routing::states::bootstrapping bootstrapping.rs:331] Bootstrapping(7c6674..) Sending BootstrapRequest to 982a7c...
D 17-08-05 23:12:53.591293 [routing::states::bootstrapping bootstrapping.rs:247] Bootstrapping(7c6674..) Timeout when trying to bootstrap against PublicId(name: 982a7c..).
Looks like the proxy after having a client disconnect from an IP refuses to accept another client from the same IP for two hours after the client has gone offline. I mention two hours as thats time the time after which the entry in that check would expire and the proxy would allow the IP again.
Ideally that check should have occurred after checking if the incoming message is a BootstrapRequest from a peer who is a Boostrapper as then that peer would be purged from dropped_clients as expected on receipt of the BootstrapRequest message but here due to the check, the BootstrapRequest message is not handled and instead just ignored, which then causes the timeout client side on waiting for a BootstrapResponse …
Just running some tests locally with a simple patch if ok might be able to update the proxy nodes with the fix. Just need to do it quite slowly between the nodes to prevent wiping the network, will update in the Dev update thread when pushing the patch and hopefully that resolves the issue.