Agreed, that is odd behavior for the loopback address on both tunnel-servers.
Probably all of the tunnel-servers have the same unreachablity.
Eg. traceroute6 tserv8.dal1.ipv6.he.net
BGP between the tunnel-servers and core routers is operational though, so tunnel traffic won't be affected by any issues with the loopback addresses.
I haven't figuered any other way of seeing if this is only tserv6.fra1.ipv6.he.net unavailable or otheres then by tracerouting (loopback) adresses.
So that was worrying that they all routed to SJC. My apologise for my comment about "all over the world".