[derefr]

I don’t think there’s such a thing as a well-implemented carousel. Maybe in theory you could have one, if browsers could do eye-tracking, such that you could stop the carousel’s rotation as soon as the user looks at it.

But without that, users are always going to land their eyes on the carousel and start reading an item in it some random interval after the carousel’s rotation timer last fired. (As, remember, users like to look away from the browser while pages load, and it takes them a random amount of time to look back.) Then, with this random "misalignment" of engagement times, the carousel content will inevitably transition out from under them.

IMHO, this is exactly why users choose to not engage with content in carousels (according to the studies the author cited.) They know they’re going to not get to fully read whatever it is the carousel currently says; and they don’t know for sure how to stop the carousel from progressing; so they don’t bother with it.

[koheripbal]

If there's no such thing as a well-implemented carousel, then the demonstration shouldn't have needed to use an obviously nerf'd carousel.

[setr]

Its easiest to point out an issue by first pushing it to the extreme, and then re-evaluating whether the middle-ground does not exhibit the flaw (in lighter fashion)