We have some host with multible JVM and java applications. we defined multible JMX connectors for each VM and created templates and linked items
to correct connector for polling.

when 1 JVM is powerdown , for example during a migration period which took 3 days , the JMX host becomes unavailible after some failed polling on this JVM.
this causes the polling of the second JMX connector to also stop and slow down because of unavailable JMX host.
items with 20 seconds poll time , will get polled but in random 15..20 minutes.
So any monitoring allerting of second item is lost. Triggers based on nodata() are trigger will the second application is still working

i would suspec that the unavailable handling would be limited per connector, and in case of JMX it should not be host global.