Bug where the composite is nesting too deep in the parallel version? The call to NeuronModelArr at lines 362 (and presumably 326) throws an error about accessing composites.
When debugging seems that it isn't accessing the nested neuron model correctly. This is triggered when trying to stop or start field stimulation in the code.