It's just a very new and therefore unknown package. We were suprised about the benchmarkings too. Turns out some libs benchmark their memoization by feeding the

It's just a very new and therefore unknown package. We were suprised about the benchmarkings too. Turns out some libs benchmark their memoization by feeding the memoized functions with one and the same argument millions of times. If you hit the memoized functions with a set of say 1000 different arguments in a loop it turns out some implementations get slow (maybe some O(n) algorithms under the hood where there could have been O(1)).