LF EQS solver diverged/failed

I ran a batch of simulations at multiple frequencies. All other settings are the same (except for the materials which are automatically updated at each frequency). Some simulation succeeded, while some failed, with different types of error: some diverged and the solver stopped before max iterations are reached, and some did not converged below the tolerance so the solver stopped after reaching max iterations. I attached some plots and log files here. Due to the size limit for uploads, I removed parts of the log files (the iterations & error). As I understand, the simulation failed because the solver met one of the stopping criteria. My question is: What can be changed about the model to have a more stable numerical problem?
sim_at_5kHz.log
sim_at_15kHz.log