Port conficts will occur when multiple pods dispatched to the same node under hostnetwork.

**Hostnetwork** needs to be enabled to utilize RDMA for high performance transmission. In such circumstance, there will be port conflicts when multiple pods dispatched to the same node for the following reasons:

- sshd of all workers will listen at port 22
- ssh will use default port 22 for connection

Can mpi-operator handles the port conflicts under hostnetwork?

The issue names no files, tests, or entry points. Start by tracing how hostNetwork pods and worker SSH ports are configured, then define and verify a conflict-free behavior when multiple workers share a node.

Written by the indexing model from the issue text.