In your opinion, is the correlation between the length of the ciphertext and the decrypted text (even if it is approximate) a cipher vulnerability? Or is there a solution to this problem outside of it?
It may be a weakness; however the solution is often too expensive to use, so we often live with this weakness.
The obvious way to address this is to pad all ciphertexts to a fixed length. Now, if we want to be able to encrypt 1 Megabyte random plaintexts, then obviously the ciphertext must be at least 1 Megabyte. Now, if we pad all ciphertexts to 1 Megabyte, that means that even a 1 byte plaintext encrypts to that 1 Megabyte.
This solution works; however in practice, it is often considered too costly (what "costly" means in this case depends on how we store or transport the ciphertext). What we can do which isn't nearly as costly is reduce the amount of size information the adversary gets (rather than eliminate it entirely). For example, if we always pad the ciphertext to the next multiple of 256 [1], then the adversary is able to deduce the approximate size of the plaintext, but not the exact value. This still leaks some information, but not as much (and is far cheaper to implement).
[1]: Alternatively, add a small random amount of padding.
