Declaring weights in simulation combiner from array changes wrong channels

I want to use the Python API to generate different predefined configurations of eight channels in a multiport simulation. As a simplified example, I would like to make eight different simulation combiners, with for every nth simulation, the power of the nth channel being 1W (and all other channels 0W). I would do this using the following code:

def multiport_SAR_export(powers, phases, config_nr):
    # Adding a new EmMultiPortSimulationCombiner
    inputs = [em_multi_port_simulation_extractor.Outputs["EM - Source 1  (pTx coil/Channel 1)"], em_multi_port_simulation_extractor.Outputs["EM - Source 2  (pTx coil/Channel 2)"], em_multi_port_simulation_extractor.Outputs["EM - Source 3  (pTx coil/Channel 3)"], em_multi_port_simulation_extractor.Outputs["EM - Source 4  (pTx coil/Channel 4)"], em_multi_port_simulation_extractor.Outputs["EM - Source 5  (pTx coil/Channel 5)"], em_multi_port_simulation_extractor.Outputs["EM - Source 6  (pTx coil/Channel 6)"], em_multi_port_simulation_extractor.Outputs["EM - Source 7  (pTx coil/Channel 7)"], em_multi_port_simulation_extractor.Outputs["EM - Source 8  (pTx coil/Channel 8)"]]
    em_multi_port_simulation_combiner = analysis.extractors.EmMultiPortSimulationCombiner(inputs=inputs)
    em_multi_port_simulation_combiner.UpdateAttributes()
    weights = em_multi_port_simulation_combiner.GetChannelWeights()
	
    for n, channel in enumerate(weights):
    	weights[channel] = powers[n], phases[n]
    
    for channel in weights: 
		em_multi_port_simulation_combiner.SetChannelWeight(channel, *weights[channel])
    
    document.AllAlgorithms.Add(em_multi_port_simulation_combiner)
    
    # Adding a new EmMultiPortSensorCombiner
    inputs = [em_multi_port_simulation_combiner.Outputs["Overall Field"]]
    em_multi_port_sensor_combiner = analysis.extractors.EmMultiPortSensorCombiner(inputs=inputs)
    
    em_multi_port_sensor_combiner.Name = "Overall Field (combined)"
    em_multi_port_sensor_combiner.UpdateAttributes()
    document.AllAlgorithms.Add(em_multi_port_sensor_combiner)

configs_powers = np.identity(8)
for config_nr in range(8):
    multiport_SAR_export(configs_powers[config_nr,:], np.zeros(8), config_nr)

As expected, I get eight simulation combiners, in which each one has a single channel with power 1. However, it is not always the case that the nth channel has a power of 1W for the nth simulation combiner. For example, I've had a case with the following combination of simulation numbers and powered channels:
simulation 1 = channel 1
simulation 2 = channel 7
simulation 3 = channel 3
simulation 4 = channel 8
simulation 5 = channel 2
simulation 6 = channel 6
simulation 7 = channel 4
simulation 8 = channel 5

This "remapping" is consistent throughout simulations: if I would in the next simulation try to power channels 4 and 5, in the simulation combiner channels 8 and 2 would be powered. However, after a restart of the computer (and changing a few things to the code) a new remapping has appeared:
simulation 1 = channel 4
simulation 2 = channel 5
simulation 3 = channel 3
simulation 4 = channel 6
simulation 5 = channel 1
simulation 6 = channel 2
simulation 7 = channel 8
simulation 8 = channel 7

Does anyone know what could cause those seemingly random changes in channel declarations? I've something is not clear, please ask me and I am happy to try explaining it better.

The problem has to do with (to me) unlogical ordering of the 'weights'-dict. I have made a workaround which extracts the values from the dict in alphabetical order of the keys, which seems to work.