Any thoughts on delimited continuations with respect to performance? Why aren't more people/languages using them?
0: https://docs.racket-lang.org/more/index.html section 11, "Continuations"
1: http://pllab.is.ocha.ac.jp/~asai/cw2011tutorial/main-e.pdf
As for why delimited continuations are not more popular, many people find shift/reset a bit difficult to program with. In particular the type systems for them are a bit odd and some variants like prompt/control don't have nice type systems for them. Currently, the closely related notion of (algebraic) effect handlers is quite popular in the functional language design community as something quite similar in expressive power but more intuitive for programming and with very natural typing. The Koka language has a lot of nice introductory resources if you are interested in learning more: https://github.com/koka-lang/koka . There's even a serious proposal for adding something based on these to webassembly: https://github.com/effect-handlers/wasm-effect-handlers .
Using CPS with closures of dynamic extent (i.e., allocated on the stack, either explicitly, or because that's all you get, or because the compiler is smart enough to figure out that these closures don't leak out) is a lot more like delimited continuations (but easier to use).