This is more of an issue now that we have congestion control mechanisms
in place where transactions can get cancelled after consensus. We also
have not run our benchmark suite routinely to push these more congested
scenarios so we have not run into this issue before,