* It is not given that the "Von Neumann Bottleneck" is what generates flabby, verbose programs.

* Computers execute programs, human reason about them. There's no reason to expect the same tools to apply. This has similarities to Linus' objections to C++ in the Linux kernel - kernel code is an implement of high level concepts and there's no to have the language structures partly support or impose these high level concepts; instead one should understand abstractly and then bring them to implementation.

* The operations of functional programming aren't efficient in their naive implementation. They can be made efficient with an optimizing compiler but this effectively results in the programmer thinking about more, not less, elements.

* Dijkstra takes issue with Backus' assumption that functional programs will make program proving accessible to the average programmer. I think Dijkstra wants to up the skill of the average programmer rather than expecting that functional programs will make proving easier. (this one I'm less sure of but it seems like you have any situation where only "unaverage" programmers are going to be using functional languages).

Dijkstra closes: "In short, the article is a progress report on a valid research effort but suffers badly from aggressive overselling of its significance, long before convincing result have been reached."

And I think the situation now is that FP has shown itself to be definitely useful in some domains but still suffers from publicity that implies it is a paradigm that will "eat" ordinary software engineering.

That it has huge mind-share for what it isn't might be as bad as it not being widely used for what it is useful for.