it doesn't seem to me like this really solves the problem, if we assume the attacker has control of the relay that was sent the channel inits, maybe i'm missing something?
also, the channel state has to be stored by both ends somehow, or it only works, like simplex, per client key
the worst a relay can do is drop messages.
the key generated for channels is completely unrelated to the init message, and it can be queried from a dedicated connection on say a different tor circuit.
maybe you haven't actually written any relay code, because you would then understand the relay sees the IP addresses they are coming from
piece of cake to just log all that and filter for correlations between channel requests and IP addresses
or, like i say, am i missing something?
I just said use a dedicated tor circuit to query for events on each channel.
You can’t correlate two different circuits via IP because they have different exit nodes.
