This happens since the submitted job for the autodetection will respawn a job if it does not find the remote host topology which will byitself submit another job.
login-node -> autodetection job 1 on compute node
autodetection job 1 on compute node -> autodetection job 2 on compute node
.
.
autodetection job n-1 on compute node -> autodetection job n on compute node