Guard, Stable, and HSDir flags don't get reset for Sybil relays
If you run ten relays on a single IP address (like is happening right now -- see also #8145 (moved)), the directory authorities will only pick two of them for the consensus.
Problem is, the authority will track uptime for all ten of them, which influences whether they get the Guard, Stable, or HSDir flag when they come back into the consensus.
We should mark a Sybil as a 'downtime' event when we decide it doesn't deserve the Running flag.