I've been using frontend in conjuction with pdsolve to handle a very large system of PDEs; namely the system has many irrelevant functions that can be frozen with frontend, allowing pdsolve to solve it much quicker.
However, I ran into a strange case, which I've included in the attached file. Namely, for this particular system of equations, the system is a PDE with respect to t, so I freeze functions of x. However, doing so produces incorrect results (labelled bad) compared to the unfrozen case (labelled good). The frontend command is working as expected (this can be verified by uncommenting the frontend/print commands); it seems that pdsolve is treating the frozen and unfrozen cases differently.
Any idea why this behavior is occuring? (The obvious solution is to simply not use frontend here; again, I'm using this same code for much larger PDEs and frontend speeds up exceution by an order of magnitude).