Comment 26 for bug 1080765

Sorry for the delay. The most reproducible (aka most frequent) case happens in a particular master/slave relationship where we are only replicating one database from master to slave. That is, the following is present in /etc/my.cnf on the slave:

replicate-do-db=vsp_jira_current

There aren't any pt-table-checksum replication filters in play, though I do only check the one database; the command line is:
${PERCONA_DIRECTORY}/pt-table-checksum-bug-1080765 --no-check-binlog-format --user=$USER --password=$PASS --chunk-size-limit=50.0 --replicate-check --pid=pt-table-checksum_ssodb03.pid --recursion-method=DSN=D=percona_ssodb03,t=dsns --databases=vsp_jira_current --no-check-replication-filters --replicate=percona_ssodb03.checksums

Within this database, all tables are InnoDB. For testing purposes, I am not clearing the percona_ssodb03.checksums table.

As to your last question:

Are any other programs running? And are you *really* sure about all that? :-)

I am absolutely sure there ARE other programs running. We don't shut down our web apps whenever we run pt-table-checksum; it's certainly possible for the apps to be hitting this database with DML. However, we have another slave that is replicating ALL databases from this particular master, and it's not seeing the problem.

Unfortunately, these DBs contain private info, so I cannot grant outside access to them. But I'm happy to help out wherever I can (understanding that it'll slow down the process).