If 800km impacts go asymptotic, it pollutes 700km and 900km orbits by virtue of having a distribution of resulting debris velocity vectors, and as drag pulls down all the resulting debris over the next thousand years, the 800km debris becomes circular 700km debris, and then circular 600km debris, and then circular 500km debris.

> as drag pulls down all the resulting debris over the next thousand years, the 800km debris becomes circular 700km debris, and then circular 600km debris, and then circular 500km debris

Circularisation isn’t the unexpected part. Sphericalisation is. One requires orbits to desync. The other requires plane changes.

There is percession of the perigee.

Even in a purely planar distribution, nodal precession still occurs slowly.

It doesn't even need to be factored in, though, if different planes are colliding with each other and energetically generating a spectrum of new orbital vectors (many less than circular) from impact. This effect colludes with altitude drop from orbital decay and the tendency to circularize orbits by perigee drag, to make it so that higher orbit debris percolate into lower orbits over time.

  > plane change

If you watch the animations in this (excellent) ESA video, you see the plane change occurs rapidly all by itself. Over the course of a few months it covers the entire globe, spreading across all "latitudes" (aka RAANs).

Roughly 3 minutes in: https://www.youtube.com/watch?v=9cd0-4qOvb0

https://en.wikipedia.org/wiki/Nodal_precession