But what if finding an answer to a problem depends on connecting seemingly unrelated atomic facts? If they are both in your brain, you may be able to figure it out. But if not, you're out of luck (unless someone has solved that exact problem before and posted it on the internet).
Science students should be getting exactly this kind of problem on a regular basis in the form of exercises, and simultaneously given the resources to dig up those facts (e.g., Google) in order to solve the problem. Then they'll learn the useful interplay among simple facts that forms the basis of relational and functional knowledge. Sitting them down and telling them to memorize the periodic table won't accomplish this.
Yeah, you have to know stuff, but simply knowing lots of facts that you're not in the process of actively using is sterile. Yeah, you don't necessarily know ahead of time which fact will end up being useful, and you might have to go out and learn new stuff to solve your problem. And a good basis of broad knowledge is really helpful for that. But that's not an argument for just making students memorize shit because I Say So.