[chatter_concert] showing symptoms when a redis server zombifies

[stonier]

Looks like there's something triggering a redis server zombie...This causes the next run of a concert to fail with no known (as yet) road to recovery. When the concert starts, you get the following problem:

On the concert:

[INFO] [WallTime: 1426556552.805717] Conductor : no hub yet available, spinning...
[INFO] [WallTime: 1426556552.805904] Conductor : no hub yet available, spinning...
[INFO] [WallTime: 1426556552.806029] Gateway : Found existing mismatched public key on the hub. Requesting resend for all flip-ins.
[INFO] [WallTime: 1426556552.806554] Conductor : no hub yet available, spinning...
[INFO] [WallTime: 1426556552.806594] Gateway : registering on the hub [Chatter Concert]
[INFO] [WallTime: 1426556552.806756] Conductor : no hub yet available, spinning...

On the client:

[INFO] [WallTime: 1426556010.183869] Gateway : discovered hub directly [http://localhost:6380]

Three issues here:

1. What triggers a redis server zombie?
2. Why are gateways providing erroneous diagnostics about the problem?
3. Can it be automagically recovered?

[stonier]

Redis server is perhaps zombifying from an untimely and early rocon_launch shutdown?