Problem:
If one adds SHOW PROCESSLIST to the end of rpl.rpl_mts_debug.test, it will show
an orphaned SQL worker thread that is there regardless that the slave is
stopped. It is a thread created by mts_worker_thread_fails failure injection
point testcase, which tries to set up MTS with two worker threads and
simulates a pthread_create failure for one of them.

Analysis:
The root cause is that the variable slave_parallel_workers will not be
incremented in case of failure in slave_start_single_worker
(caused here by "mts_worker_thread_fails" injection). This will end up
causing an early exist in slave_stop_worker

void slave_stop_workers(Relay_log_info rli, bool mts_inited)
{
  int i;
  THD thd= rli->info_thd; if (!mts_inited)
    return;
  else if (rli->slave_parallel_workers == 0)
    goto end;

Fix:
Incremenet the variable slave_parallel_worker for every successful call to
the method slave_start_single_worker(). In this way the early exist will be
avoided when there are worker threads present which needed to be stopped.