Work that runs in the background has an awkward property: you do not watch it. As long as everything goes well, nobody notices anything. When something is stuck, nobody notices that either, until somebody complains about an email that never arrived.
Horizon closes that gap. The interface shows how many jobs are waiting, how long they have been waiting, which have failed and on what. A failed job can be started again with one click once the cause is fixed.
Why this is not a toy
The real value is in the numbers over time. When a queue is normally empty within seconds and today takes forty minutes, that is a signal long before anyone phones. Without numbers like these, all that is left during an incident is guesswork.
Two things need settling. Horizon assumes Redis underneath, which has to be allowed for in the planning. And access needs restricting, because there is real data in the jobs. Both belong, for us, in the operating agreement under maintenance and support.