Possible questions of interest:

What are the limits of multi-unit recording? How fast are these limits changing?

Is there a best way to sort spikes? Or rather is it application dependent, and if so what are the parameters of these applications? Is there a best method to detect structure in multiple unit recording? More generally, is there a common set of problems faced by all researchers who do multi-unit recording?

Do we know what and how computations are performed by nervous systems with relatively few neurons? What can spike data from every neuron tell us?

What are the goals of multi-unit recording? Do we want spike data from every neuron in a cortical column? What are the best model systems for getting this data?