E2: transverse estimates, explicitly unfinished

E1 has not been crossed. No scaling degree is assigned here to an undefined full triangle or four-vertex distribution.

For a collision subset $I$,

Choose local coordinates $(q,r)$, with $q$ tangential to ${\Delta }_I$ and $r\in {\mathbb{R}}^{q_I}$ normal to it. A prospective off-diagonal distribution $a(q,r)$ would be rescaled by

In distributional notation this includes the usual ${\rho }^{-q_I}$ test-function Jacobian.

The required estimate is not merely a bound for one fixed smooth value of $q$. It must hold uniformly on compact tangential sets, with the relevant distributional and wavefront seminorms controlled, including the approach to the other collision strata.