An interesting development is that the argument "common libraries will be cached in the browser" is no longer true. Chrome and other browsers are starting to scope their caches by domain, to mitigate tracking techniques that used 304 request timing to identify if the client had visited arbitrary URLs.

Yes, I'm aware that "it will be cached" lost most of its glory when bundling became mainstream, but I still hear it as an argument when pulling things from common CDNs.