With all due respect, Andy, Atlassian's position seems to have a very narrow view of the purpose of story points. Story points have nothing to do with speed. They represent the size of the work, which is comprised of effort, complexity and risk. As pointed out in an excellent article on DZone (https://dzone.com/articles/is-there-any-value-in-estimating-if-youre-using-ka), the size of the work is used to help guide prioritization and splitting.
Prioritization is still a thing in Kanban. Constant change merely means constant changes in priority as well. Without the size of the work, prioritization has a serious blind spot in its decision-making.
As for splitting, Kanban boards are fairly useless unless items are moving regularly from one column to another. When an item sits like a lump in the same column for weeks on end, transparency is lost and executives start to question the value of Agile. I've seen teams abandon the use of standups entirely because the status is the same every day. Work that is very large must be split up into smaller tasks that can be tracked. Sometimes that is done with sub-tasks and sometimes it needs to be done at the Story / Task level. Without sizing, that work cannot be done, putting Agile at risk.
Finally, there is the issue of cycle time. I think it should be obvious that the cycle time of a 3-point story is different than the cycle time of an 8-point story. A single cycle time metric that integrates all sizes is horribly inaccurate and cannot be used as a predictor for any given story. Cycle time should be calculated from stories / tasks that have the same size. That would make more sense.
Sizing is important. It's a thing. Kanban needs it. If you don't want to use story points, what do you propose to replace it?