Yep, it is still a matter of abandoning the monolith and building a new scalable system in whatever fashion you choose at that time, micro services or whatever.

In today's cloud deployments where cost is not a mega concern (as compared to having physical servers in a DC), this approach is very much feasible.

You could run two systems in parallel (monolith v/s scalable) and ditch the monolith once you have the necessary cutover logistics in place.

Of course, the sooner you do this, the better and it is non trivial to figure out the 'when'