This is to reproduce crashes with mariadb 10.3

I believe that in the live system where we observe the crash we have expand
taking more than 5 seconds, and that if we use a shorter timeout here the crash
happens more often.