Evolution is whatever survives and reproduces lol. I think the good point of this model is that not necessarily all the neurons spike at the same time. Personally, i think that sparsity is help the network to train better, allowing to better generalization since only the relevant patterns are learnt. I see spike NN as a differential sparsity (my opinion and i stand by the consequences if I say something stupid), that force the netwrok to learn pattern that are needed in certain condition but not in other, so the sparsity is conditional.

The last author is known for the work on spike NN. It is not very easy to work with spike NN, especially because default backpropagation and gradient descent are not actually working, so you have to modify.

Then there are many variants of spike NN each claiming to be the best theoretical framework. I do not think there is a benchmark