slave will skip the first SQL_SLAVE_SKIP_COUNTER events on start - 
this is needed to recover from broken replication when the next log
event(s) should just be skipped.
Rotate binlogs when max_binlog_size is reached