• There are lots of bad keys, of the form A^n/B^m, where A, B, n, and m are integers.
  • A known plaintext recovers the keystream. With a little work, it could possibly recover the key, but only for very long plaintexts.
  • The key cannot be re-used. Given 2 ciphertexts encrypted with the same key, add them (mod 6 digitwise). This gives plaintext 1 + plaintext 2. It's pretty easy to recover the texts from this - there aren't that many possibile ways for 2 meaningful English text streams to be combined to form a given stream.
  • Similar (numerically almost equal) keys will give streams with the same set of initial digits. This can be trivially fixed by simply ignoring the first 15 or so digits.