A bridge's new identity is not actually ignored although we say so
I have got a bunch of these warnings recently since my bridge had a full reinstallation.
entry_guard_learned_bridge_identity(): Bug: We 'learned' an identity [redacted] for a bridge at [redacted], but we already knew a different one ([redacted]). Ignoring the new info as possibly bogus. (on Tor 0.4.0.1-alpha 81f1b89efc94723f)
But it is not actually ignored: Tor bootstraps as usual, builds circuits using its new descriptor and writes circuit building state under its new identity. This could create an MiTM possibility, or maybe we need to elaborate a bit more on this.