I was setting up a slave and doing some table checksums to check things were ok before getting them live.
I did a native mysql checksum tables on the slaves after syncing slaving to a stopped point and comparing against a preexisting slave.
Two tables differed. To drill down I ran pt-table-checksum which reported both the old and new slave had no differences and the same row counts.
To get a third opinion I did a select * into outfile ordered by PK and ran an md5sum on the outputs. Those too both matched.
One noticeable difference is the newer slave is running a slighter newer version (5.5.34 vs 5.5.29); both percona builds.
Is this version difference something to be expected in causing differences in checksums? If so it seems odd it was just two tables out have a couple dozen.