What about the witness log?

It looks like the mirroring was broken a few minutes before the principal reboot, that explains why there was no failover when the principal rebooted (mirroring already dead), but not why it did not failover on network failure.

Now if for some reason the principal lost connection with the mirror and the witness at the same time, the mirroring would be broken from the Principal point of view.

If the mirror lost connection with the witness and the Principal, the mirroring would also be broken from it's point of view.

If something happen to the principal after that, there would be no failover.

Other solution is that you have no witness, then there would be no automatic failover 🙂