This is highly dependent on the culture of the specific site.
For example, the idea of a sandbox actually originated from Programming Puzzles & Code Golf. It's highly successful there because questions on PPCG are so different from other Stack Exchange sites. Questions are posted not to request help but to pose a challenge, and they typically require much more work and fine-tuning than questions on other sites. There are also lots of ways to go wrong in writing a programming challenge (loopholes, etc.), so a sandbox is useful to weed those out.
Worldbuilding is also a much broader and more open-ended site, so it possibly makes sense to have a sandbox there, too. (It's not as "special" as PPCG, though, so a sandbox isn't as necessary in my opinion, and indeed it's much less active than PPCG's.)
The question you should really be asking yourself if you're considering whether a site needs a sandbox is: What problem would this solve? In the case of PPCG, as @ChrisJester-Young puts it, "Currently, the sandbox is a buffer against duplicate, poorly-specified, uninteresting, or just generally inviable challenges. It makes the task of framing challenges a collaborative, community-orientated effort." For Worldbuilding, it's a way to help new users out and ease them in to the slightly unorthodox site and community, compared to other sites. It's really just a thing that should be decided on a case-by-case basis.