A lot of good experience has been written for finding problem queries on this thread. The experiences can be summed up by saying that for a person to identify potential problem queries, one should focus on queries that are consuming one or more server resources at a distinguishing level.
Gila Monster's post regarding high overall CPU utilization for a short executing query was a particularly useful and pertinent example.
But, before we move on to another post and another day, I am going to revisit long executing queries.
Long executing queries “occupy” a server resource: Time. Unless a query is executing a "WAITFOR DELAY.." command, if it's executing, it's either doing something or having something done to it. It's either consuming a server resource(s) or it's being blocked. In either case, long executing queries are candidates for investigation, to learn what resources they are consuming and how they are doing it, or, if they are being blocked, what is extending their execution time? That is, “What is (are) the blocking point(s) and how can they be eliminated or minimized?”