nmis does collect and display a number of configurable health metrics that cover a device's recent past. these are documented on the NMIS Key Performance Indicators page, and they include the "interface availability" of a node, tracking how often and how long an interface has been down.
to react to chronic repeat problems of that kind you can use a threshold alert, and nmis comes with a standard threshold alert for interface availability.
at this time nmis does not have a mechanism for reacting to a particular number of occurrences of a problem in a time window, but our event management application opEvents does support this - see Event Correlation for details.
Wondering if NMIS has the ability to identify a chronic problem. For example, an interface that has gone down 3 or more times in the last 72 hours.