Metricbeats output.redis excessiv DNS Querries when redis down

We have metricbeats sending data to redis and on to ES. When redis is down metricbeats CPU jumps to 100% and 4 hosts easily create 20,000 DNS requests/second for the list of hosts running redis. Kinda odd it's not cached at all.

Any way to address this? Kind makes it unsafe in our lab.

#-------------------------- Redis output -------------------------------
output.redis:
    hosts: [ "<list of redis hosts>" ]
    key: "metricbeat"
    db: 0
    timeout: 5