Familiarity is always something I take into account when choosing tech for a new project, but I also balance that with other factors like speed of development, is this a very long-term project or something that needs to be done yesterday, how mature the tech is, etc
If it needs to be done "yesterday", then choosing something you're not familiar with is almost setting yourself up for failure, I've seen many projects fail or go waaaaay over budget simply because developers weren't familiar with the stack and in cases where it was a complicated stack, very buggy code emerged. On the flipside, I've seen many more projects succeed where developers were familiar with the stack, but there were still failures in some projects due to other problems - long story short, familiarity with a tech stack will increase your probability of successfully completing a project.
If it's a R&D type of project or something which doesn't really have an immediate deadline, having the flexibility to play with new tech is sometimes fun, but you will often churn through many months of trial and error before you get to a point where you fully grasp the technology and are able to start building useful stuff with it.
I've also been burned in the past by tech that's not yet fully mature - often you'd get stuck and nobody would be able to help you. Many years ago when GWT was the new kid on the block, you would post on StackOverflow and not get an answer at all, post something about jQuery and by the time you hit submit and refresh, you already had two answers. So something I also take into account is maturity, how many well-maintained libraries are available, what kind of support are available for it - do you have an army of people waiting to respond to your questions or are you the lone ranger battling through language / framework quirks all by yourself.
Also on maturity, is the language / framework about to undergo major changes that are not backwards compatible that will break your system? If this happens often, I would often disregard the framework / language as a viable option.