Nonconvergence of radiation RTE equation in coaxial flame and setting of radiation boundary conditions

[grabtheNS]

Hello, everyone.
I encountered an MLMG solver solving more than the number of max iterations [1000 Max iterations] (also encountered divergence cases)
I'm sure the problem comes from solving the radiation RTE equation, i have try to https://amrex-combustion.github.io/PeleLMeX/manual/html/Troubleshooting.html ，and reduce the time step (as low as 2E-6)But it doesn't seem to work. my case is similiar to Exec/Production/DiffBunsen2D

peleLM.lo_bc = Symmetry      Inflow
peleLM.hi_bc = SlipWallAdiab Outflow

I suspect that it is the problem of boundary conditions. I have consulted the code of the radiation part, and it seems that only the fixed boundary conditions of robin type (fixed parameters) can be selected for radiation boundary conditions. This may conflict with my model boundary conditions (which I have symmetric boundary conditions on one side),
https://github.com/AMReX-Combustion/PelePhysics/blob/development/Source/Radiation/PeleLMRad.H row266

      if (robin_cell) {
        robin_a_fab(i, j, k) = -1.0 / betafab(i, j, k);
        robin_b_fab(i, j, k) = -2.0 / 3.0;
        robin_f_fab(i, j, k) = 0.0;
      }

But I can't locate the exact source of the problem, before the radiation calculation, my flame (with soot) has achieved steady state, if someone has a similar problem, we can communicate and solve it

[baperry2]

@wjge do you have any suggestions on what might be happening here or how the boundary conditions should be set up in this case?