One point there really resonantes with me as someone who reads a fair number of tech books.

"5. Choose a topic that will sell in a year and half."

makes it hard (I think) for books to be written about very fast moving technical topics (e.g. Docker/Kubernetes/CoreOS). By the time you've written the book, it will inevitably be out of date.

The approach that publishers like Manning and O'Reilly take of publishing beta books helps a bit but still these topics are not a recipe for long-term saleability.