Since nowadays there are plenty of statistical software, it is not essential to have a deep background knowledge of statistical concepts (but I say it is important). And I think its depend upon the research objectives, if it is a some common analysis problem which does not include any mathematical concept you will not need any expertise knowledge in pure mathematics but if the your research is theoretical one you will happy to see the joy of knowing pure mathematics. Also it is better to have some confirmed understanding of mathematics to follow research papers. So I think if you are a full time researcher in statistics you better have a knowledge of pure mathematics.
If you do not know anything about the central limit theorem and convergence of random variables, how do you assess whether you can use asymptotic tests for goodness of fit? And this is just an example.
My reply to the original question is: If you work in statistics, the more you know (on probability theory, measure theory, etc.) the better.