> The expected lifetime value of any article in Nature Neuroscience for the authors is well above $11,390

This. Authors are the customers of the journals; the service they are buying is a push towards tenure and more funding. This has been true for a long time; open access only makes it more explicit.

When readers are the ones paying, you can argue that they are paying the journal so that it makes sure that they are reading serious, properly reviewed articles.

When authors are the ones paying, how do incentives align?

I think journals are now in a situation similar to the one of rating agencies and accounting firms. I don't think it will work out very well.