That was actually my main personal discovery in the process of developing Leibniz. I started out with a first implementation in Racket. I had a Scheme background, and Racket has a lot of support for developing languages, so it seemed like a good choice. And if you consider development tools a separate subject, as most of us do, it is. Implementing my first design of Leibniz in Racket (using Emacs as my IDE) was a good experience overall. But improving on the initial design, iteratively, turned out to be very frustrating. I had two nested edit-compile-run cycles, one for Racket and one for Leibniz, and no other tools than print statements for looking at the internal data structures of my implementation. That was way too cumbersome to handle. When I discovered Pharo (by accident), I immediately saw that a live environment is what I wanted. So I started my second implementation in Pharo, which at that time contained a precursor of today's Glamorous Toolkit that was already very useful. Since then it only got better.
My current point of view is: you can't design a language without in parallel developing an IDE for that language. And that becomes an important criterion for choosing a platform.