1. OCX on H1 acts on R1 because its repository matches the one named in the run. The other host's OCX does nothing with it.
2. OCX on H3 does not act on R1 because its default user differs from the one named in the run.
3. The `ended` message counts because it comes from the repository's default user, names the correct repository and is not sent by the run itself.
4. OCX on H1 acts on R2 but zed's `started` does not count because the process user "zed" differs from the run's specified user "builder".
5. A run could name `"process": {"user": "lab-1"}` and be run if lab-1 holds a capability that allows it.
6. The most OCX on H1 can do on the server is set by who has the repository's command line, chosen by either an admin or setup user.
7. `$$clone$$` refers to `$workspace$/clone`. It is not the same as the repository named in the run's `repository`.
8. A new run naming the old server-code aaaa…a arrives and is answered by no one because clones have unique server-codes.
9. A run names `default_user` instead of OCX finding the user from the server-code to ensure that only an authorized user can report on the run's success or failure, preventing unauthorized users from altering results.
10. Anyone who holds admin or setup may send a `stop` for R1 and it holds the keys `ocx`, `kind`, and `run`.