There is a period during the bootstrapping process (when there are still very few verified servers in resolver_q) and possibly near the end (when the list is exhausted) when a resolver might only have one nameserver. In this case, it will hammer that same resolver query after query with no break. This change puts in an explicit wait for these transient times in order to keep to the implicit contract of not querying the same server more than about once every 5 seconds.