Description of problem:

When running a REX job on multiple machines, with concurrency set to less than that number of machines, there are some tasks that are in planned state - they are not running yet. When I select all the tasks and under kebab menu, "Cancel selected", a success is shown (with a number of tasks that has been cancelled, minus the ones planned) and the planned tasks continue running and finish.

How reproducible:

Deterministic

Is this issue a regression from an earlier version:

Reproduced on Sat 6.18 snap 6.0. Happens in new REX UI, this action wasn't possible in the legacy UI (only cancelling the whole job).

Steps to Reproduce:

1. Have a Satellite with 4 registered hosts

2. In Hosts, select these hosts and click "Schedule a job"

3. On the second tab, enter some script that takes time to run, like `echo $(date) > /tmp/test; sleep 60; echo done >> /tmp/test;`

4. On the third tab, set concurrency to 2

5. Submit

6. Select all the tasks and in the kebab menu, "Cancel selected"

Actual behavior:
Green "success" message shown: "2 task(s) successfully cancelled", and the two planned tasks actually started and finished.

Expected behavior:
The running tasks AND the pending tasks should be cancelled (that means the running tasks are stopped and the pending tasks are not started, even in the future).