The init script was waiting for cloudbr0 to come up, but it's not mandatory that this bridge is available.
We now wait for at least one bridge to be up before starting the cloud agent.
Notice, even after this fix, there would be 2 entries in the mshost table, after
mgmt server restarted. But it should be harmless, since only one of them is
"Up".
Some message in the log would indicate something wrong with the original mshost
entry, but still, it seems harmless.