1) Index usage DMV's are emptied when SQL Server is restarted http://msdn.microsoft.com/en-us/library/ms188755.aspx

2) When properly, precisely done checksums (all columns fixed width or very, very careful delimiting (CHAR(31), perhaps), COALESCE/ISNULLed to an impossible value, etc.) can tell you:

2a) If two rows are different

2b) If you need to check further into whether two rows are the same or not; regardless of the theoretical collision chances based on statistical assumptions, I've seen many reports of actual collisions, particularly since real (unencrypted) data very rarely looks random; you can get "Bobs" "cars" and "Bob" "scars" easily enough (see prior comments about properly and precisely done).

3) Lastupdate type columns yields false "the data changed" results if a table was rebuild even when all data except the lastupdate column is 100% identical.

4) For full comparisons, I prefer something like:

WHERE (

((Old.[field1] <> New.[field1]) OR ((Old.[field1] IS NULL) AND (New.[field1] IS NOT NULL)) OR ((Old.[field1] IS NOT NULL) AND (New.[field1] IS NULL)))

OR ((Old.[field2] <> New.[field2]) OR ((Old.[field2] IS NULL) AND (New.[field2] IS NOT NULL)) OR ((Old.[field2] IS NOT NULL) AND (New.[field2] IS NULL)))

)

Note that for tables with many columns, having an automated way to build these statements is much better than copy/paste/modify.

For anyone that insists on using checksums, I would recommend a scheduled process (weekly, monthly, quarterly) that does a thorough check for collisions. Gamblers hope for good look... we should check regularly for bad luck.