We have an issue where one of the developers ran a large query that hung was filling up the DATA drive very rapidly. From 50% usage to 95% in less than 2hrs.
It created a very large pgsql_tmp size (300GB). To stop the drive filling up we had to kill the process manually using kill -9.
Did you try kill -15 first? If you did try it and it did not work, then that is probably a bug. And if you have a reproducible test case, then we will probably be able to fix it.