A pp-wave spacetime in Brinkmann coordinates has metric $$ ds^2 = H(u,x,y) \, du^2 + 2 \, du \, dv + dx^2 + dy^2 $$ and is asserted to be a Lorentzian manifold, i.e. has index 1.
This is indeed true assuming assuming that $H(u, x, y) \ne 0$. However, when $H(u, x, y) = 0$, then the metric looks like $$ ds^2 = 2 \, du \, dv + dx^2 + dy^2 $$ which has index 0, since this can be transformed to $$ ds^2 = du'^2 + dv'^2 + dx^2 + dy^2\text{.} $$
It's stated that $H$ can be any smooth function, and the Brinkmann coordinates article specifically mentions that "The coordinate vector field $\partial_u$ can be spacelike, null, or timelike at a given event in the spacetime, depending upon the sign of $H(u, x, y)$ at that event."
But doesn't this conflict with the definition of a semi-Riemannian manifold, which requires that the metric have constant index everywhere? If so, what kind of manifold exactly is a pp-wave spacetime?