I'm not sure it's helpful to ask "is that kosher?". That is, there isn't some external rule saying what sort of things are "valid" and what aren't. The questions that matter are ones like these: Is it realistic to expect this to be solved by actual human beings in finite time? Is it likely that they'll have fun doing it?
A puzzle is more likely to be impossible or not-fun if the only way to solve it is a boring search over a large space of possibilities, with no way to reduce the brute-force effort. Avoiding this for cipher-type puzzles generally means either (1) providing some extra information about what's going on (punning clues embedded in the descriptive text; information about the characters involved that may help guess what sort of cipher and what sort of keys they might have used; key information hidden steganographically in images, punctuation, etc.; and so on) or (2) providing enough ciphertext to enable approaches other than brute force (letter-frequency analysis for substitution ciphers, slightly fancier techniques for Vigenere, etc.) or (3) making the cipher "obvious" enough that it's feasible to just look at the ciphertext and figure it out. Note that #3 is probably not much fun in the first place, but it might have a place as a subsidiary element in a larger puzzle.
If you have a cipher that's made up of lots of little sub-ciphers -- let's say three, as in the first example in the question -- then #1 now needs 3x as much extra information, which may mean 3x as much apparently-irrelevant "padding", and #2 needs 3x as much ciphertext, and #3 requires you to have multiple different immediately-guessable ciphers. All of those are a bit problematic.
And if your individual ciphers are all about as difficult as a typical cipher, then you have just made your puzzle take 3x longer to solve. Or if they're easier, then you may have made your puzzle easier. You might prefer to avoid those things.
So usually this sort of combination of ciphers is going to be harder to make into an approachable and fun puzzle.
But "usually" isn't the same thing as "always", and if you find something approachable and fun that involves multiple ciphers then there's nothing wrong with that.
None of this is really specific to ciphers; it applies to any sort of puzzle. If your puzzle has multiple components, then you need to provide enough information for all of them to be solvable (which will make it longer than a single-component puzzle) and it will either take a long time to solve or be composed of fairly easy parts. These will often be problems, but none of them is a fatal objection and there can absolutely be good puzzles with multiple components.
Here's an example of a recent puzzle that has multiple components and was pretty well received: This is more than just a dream.