Duplication of data?

[rjbutterworth]

I have a StitchData integration tracking opens and clicks tables in my Campaign Monitor data, which pushes CSV files to an S3 bucket at hourly intervals.

I assumed that only new data would be pushed in these CSVs (I understand the replication method is key-based incremental), but clearly there's loads of duplication going on. As an example: in a 24 hour period, looking at the data in Campaign Monitor, I see ~800 clicks in total. However, if I merge the (69!) CSV files generated in the exact same 24 period, I end up with over 133,000 rows of data.

Even during times where there is clearly zero Campaign Monitor activity (3am - 4am on a Sunday night/Monday morning - no mailings having been sent) I get thousands and thousands of rows of data pushed to S3.

Clearly I'm doing something wrong, so would appreciate some help.