Not my field either but using (Chebyshev) polynomials to approximate _exponential functionals_ on [believably] infinite or scale-free graphs sounds rather interesting..
On further reflection, as you said, wavelets are local on space and phase-space so one has a natural cutoff..