hopbeat-monitor does not recover when a broker goes offline

[cnweaver]

A recent failure shows that if a broker disappears, hopbeat-monitor gets stuck, and must be manually restarted, even when other brokers in the cluster remain up, and the original broker becomes available again. Logs from such a failure look like this:

Received beat: 80720 1659541316522791 1659541316503461  
Received beat: 80721 1659541317531737 1659541317504838  
Received beat: 80722 1659541318542871 1659541318522128  
%6|1659541318.753|FAIL|rdkafka#consumer-1| [thrd:sasl_ssl://kb-2.dev.hop.scimma.org:9092/2]: sasl_ssl://kb-2.dev.hop.scimma.org:9092/2: Disconnected (after 177242945ms in state UP)
internal kafka error: KafkaError{code=_TRANSPORT,val=-195,str="sasl_ssl://kb-2.dev.hop.scimma.org:9092/2: Disconnected (after 177242945ms in state UP)"}
%3|1659541318.924|FAIL|rdkafka#consumer-1| [thrd:sasl_ssl://kb-2.dev.hop.scimma.org:9092/2]: sasl_ssl://kb-2.dev.hop.scimma.org:9092/2: Connect to ipv4#172.18.10.52:9092 failed: Connection refused (after 170ms in state CONNECT)
%3|1659541318.925|FAIL|rdkafka#consumer-1| [thrd:sasl_ssl://kb-2.dev.hop.scimma.org:9092/2]: sasl_ssl://kb-2.dev.hop.scimma.org:9092/2: Connect to ipv4#172.18.10.52:9092 failed: Connection refused (after 0ms in state CONNECT, 1 identical error(s) suppressed)