SELECT INTO OUTFILE query goes several times to one slave

1. execute SELECT * INTO OUTFILE '/tmp/t1.csv' FROM t1;
2. check /tmp/ directory on all nodes
3. check error log

Expected result

• error log is empty
• /tmp/ on all nodes contains t1.csv file

Actual result:

Error : Failed to execute SELECT * INTO OUTFILE '/tmp/t1.csv' FROM in 192.168.122.119:3306. HY000 File '/tmp/t1.csv' already exists
2015-07-20 21:52:44 Error : Failed to execute SELECT * INTO OUTFILE '/tmp/t1.csv' FROM in 192.168.122.119:3306. HY000 File '/tmp/t1.csv' already exists
2015-07-20 21:52:44 Error : Failed to execute SELECT * INTO OUTFILE '/tmp/t1.csv' FROM in 192.168.122.119:3306. HY000 File '/tmp/t1.csv' already exists

master has t1.csv
first and second slaves do not have t1.csv
last slave has t1.csv

Markus found that bug is not reproducible if router definition contains max_slave_connections=100%