To illustrate, consider this trust list with max-depth 3:
The system processes it by depth. So it goes through this as first (A, D, F), then (B, E, X), then (C, X), then (X).
If A
or D exclude X, then X will still have 0 net inclusions (+1 from F) and therefore be included. If they exclude F, then this doesn't help because F is at the same depth. If A
and D exclude X, then X will have -1 net inclusions and therefore be excluded. Excluding F still doesn't help/hurt.
If F changes his trust of X to distrust of X, then X has -1 net inclusions at depth 1, so his net inclusions of 1 at depth 2 and 1 and depth 3 don't matter: he's excluded at depth 1. Whether or not B, C, or E are included/excluded by anyone here doesn't matter.
If instead E and C both change their trust of X to distrust of X, then X still has 1 net inclusions at depth 1, so his net inclusions of -1 at greater depths don't matter.