Does anyone know how "depth pre-pass" is implemented?

It can make incorrect translucent renderings correct, and I want to learn this method.

Does it use a “depth peel” algorithm or something like weighted average?

hello and welcome!

we are doing several work on that front but the best way would be to start with a repro in the playground so we can all try to offer ideas

I have solved it, interested friends can look here.