In my experience the drag caused from the thinking to plan for scalability early has been so much greater than the effort to rearchitect things when and if the company becomes a unicorn that one is significantly more likely to become a unicorn if they simply focus on execution and very fast iteration and save the scalability until it’s actually needed (and they can hire a team of whomever to effect this change with their newly minted unicorn cachet).
Not only that but the work to rearchitect for scalability is often appealing, energizing work that lots of people want to do. While the work of maintaining overly complex software full of affordances for scalability scenarios (or use cases) that never come up is rarely as productive or motivating.
