I have an execution plan for a fairly complex join which shows an index seek being performed on a table with the "Actual Number of Rows" reading ~70,000, when there are in fact only ~600 rows in the table in total (the estimated number of rows is only 127).
Note that all of the statistics are up to date and the input parameters to the query are exactly the same as the parameters that were entered when the proc was compiled.
Why is the actual number of rows so high, and what does the number "Actual Number of Rows" really mean?
My only theory is that high number of rows is related to the nested loops, and that this index seek is being executed a number of times - the "Actual Number of Rows" really represents the total number of rows over all executions. If this is the case is the estimated number of rows also meant to be the total number of rows over all executions?