The fastest way to see a system stop is to stop watching the numbers

The background

Watching a message bus is the oldest debugging tool there is, and it is usually a list. It is complete, honest, and useless for the question people have: is anything still happening? A list scrolling at forty lines a second and one at four look identical from a metre away; a stopped list looks like one you are not watching.

The requirement

Answer is this moving, and which part stopped from across a room without reading a number — and do not move when nothing is arriving: a display that animates on its own asserts activity, and a pulse at a dead system is worse than a blank screen.

The options considered

A scrolling waterfall fails at once: its drift runs on a clock, so the picture moves on after the last message. Placing marks by how long ago they arrived fails more subtly: every position is a function of the current time, so it redraws on a timer. What worked was placing each mark by where it sits in the receive order, making the display a pure function of the messages received: it advances when one arrives and never otherwise.

That paid twice. Marks measure against the newest message anywhere, so stopping a producer drains its lane rightward rather than freezing it — nobody designed that. And the constraint became checkable: stop the simulation, assert the page is byte-identical thirty seconds later. Broken with a timer it failed; broken with a CSS animation it passed — motion the markup cannot see — so a second check reads the stylesheet and refuses it.

The demo

Open it at the messages tab and stop the sensors from the operator view: the observation lane drains, the control lane beats on.