[mlindner]

> At what point do they start breaking up into unrelated orbits?

It's not as easy to see as in this visualization, but Jonathan McDowell (https://twitter.com/planet4589) posts graphs on his website of each launch of starlink satellites as they raise orbits. https://planet4589.org/space/stats/star/starstats.html

Scroll down to the individual launches and there's images you can click on of the orbit raising progress of each launch.

Example of the Starlink 4-21 mission that launched on July 7th of this year: https://planet4589.org/space/stats/star/spl51.jpg (One satellite failed, which isn't that unusual, and will de-orbit probably sometime early next year if they don't recover it.)

[mncharity]

And the adjacent graphs show phase (line growing longer, spreading along an orbit), and plane (line breaking into multiple lines slotting into nearby orbits). Positioning is accomplished by spending time at different-than-deployed altitudes, mostly lower. Lower gives a difference in nodal precession (Earth being non-spherical), IIRC something vaguely like half a degree/day difference out of several deg/day, slowly changing plane westward. Thus plane changes take weeks of drifting. Launch being far more about "go fast (in some direction)" than "while being high", rapid plane change (direction change) would have prohibitive "launch-like" energy cost. So a single launch will populate one or few nearby planes. And finally lower orbits orbit faster, quickly overtaking deployed sats in a single plane to reach deployment positions. Kerbal Space Program is thought a fun way to play with such.