Let $q$ a $n$-variables indefinite quadratic form with integer coefficients, which has a non trivial isotropic vector, let say $z\in\mathbb{Z}$.
Easily this form is $\mathbb{Z}$-equivalent to a form of the shape: $$ f(x_1,\ldots, x_n) = g(x_3,\ldots, x_n)+ \sum_{i=1}^n x_2\alpha_{i}x_i $$ with $g$ having the same signature as $q$.
Is this result can be extended when in presence of a subspace on which the form is degenerated? (the previous case is then the case where the subspace is of dimension 1). I was thinking of having a family $z_1, \ldots, z_i \in\mathbb{Z}^n$ so that the form has discriminant 0 on $\langle z_1, \ldots, z_i\rangle$, and being able to find a $\mathbb{Z}$-equivalent form of the shape: $$ f(x_1,\ldots, x_n) = g(x_{i+2},\ldots, x_n)+ \sum_{i=1}^n x_{i+1}\alpha_{i}x_i $$
A partial result consists in isolating isotropic vectors in $\langle z_1, \ldots, z_i\rangle$ and apply successively the construction for one vector cited above. But it's not necessarily giving such a result in the general case.