There are two types of alarms: event-based and performance. Performance alarms can "resolve themselves" if referred value returns under the preset threshold. Event-based alarms can be automatically resolved only if you have resolve events, otherwise you need to do it manually.
I had a VM total disk latency alarm and a few minutes later it changed status to resolved while I was trying to figure out what was wrong
Response time for read and write operations changes constantly. In some period the value can exceed configured threshold, in some not, it`s a performance alarm. You can find out what caused the alarm appearance by checking related events happened at the same time.
what would be an ideal way to monitor alarms on Veeam One?
It pretty much depends on your needs and capabilities.
Veeam ONE consists Alarm Management for a comprehensive, flexible control. You can also consider the most critical alarms and set email notifications.