So, you have vacuum in the cell... In general, it will not work to determine "c" self-consistently then; instead, you will need to fix "c", and perhaps fit its value based on a reference band gap or whatever.
The warning comes from the fact that one term in MGGA diverges if the density (or rather, it's gradient) goes to zero. So while 0.03% is not a lot, it's actually more complicated than that. Note that it says the potential "diverged in 0.03 % of the simulation volume" - yes, in a small volume, but it DIVERGED! So the potential now is infinite (or, 10 H which we set it to) somewhere in space where it probably should be zero, or 1 eV, or some other value. Thus it may still cause problems.
