Combining --watch and --interval with etcd backend causes confd to miss updates

It appears that if you pass both --watch and --interval options when connected to a etcd backend, confd has issues where it can miss the second change if two updates are done rapidly. Both are running on Centos 7 servers in a distributed cluster.

Removing the --interval option fixes this, so perhaps on should just ditch the --interval option when --watch is passed. Or fail with a warning.