Boeing is going to be unable to return its 737MAX to commercial flight before 2020; the current latest guess is January, and that’s likely to slip.

Fixing the Boeing Co 737 MAX’s hazardous flight-control software and completing other steps to start carrying passengers is likely to stretch into 2020, an increasing number of government and industry officials say, even as the company strives to get its jet back into service still this year.
The situation remains fluid, no firm timeline has been established and Boeing still has to satisfy US regulators that it has answered all outstanding safety questions.
…
Boeing executives, FAA engineers, and international aviation regulators have steadily expanded their safety analyses to cover a growing list of issues spanning everything from emergency recovery procedures to potentially suspect electronic components.

That pretty much says it all.  Boeing needs to stop taking short cuts—legitimate as those might seem, and are in a variety of cases—and test this aircraft from the ground up as though every component in the control system were newly developed and never before tested.  “Every,” here, means physical components, electronic components, and software.  And pilots: their understanding of and interaction with the control systems, and with their training and operating manuals, needs to be evaluated de novo, also, as the Ethiopian crash suggests.

Too much of this stuff have not interacted with each other before in this particular aircraft, and even those components of subsystems and those subsystems themselves that have interacted within themselves have not, necessarily, interacted with other components and subsystems.

It would, in fact, be a good idea to do this from-the-ground-up retesting as a matter of routine whenever there’s a major software or hardware update to the aircraft.  Those updates create a nearly new aircraft, and accumulating changes in performance, however subtle or seemingly minor, can falsify a number of underlying assumptions and add up to large failures.

In the present case, the bit-by-bit testing is not saving much money, either.