It's not the mathematics that makes their decryption in the 1940s interesting so much as the fact that it occurred contemporaneously with the advent of general-purpose computing and information theory (which was apparently motivated by this very problem).
Usually such beginner exercises will give you >10 uses. In Cryptopals 3-19, another n-time pad exercise (labeled "moderate difficulty"), you provide n=40.
[A] http://cryptopals.com/sets/1/challenges/6
[B] "When the Russians developed their manual system (which may have been inspired by the U.S. work or a German one-time pad developed earlier in the 1920s) they presumably reckoned that using them twice was safe enough.
"They were spectacularly wrong."
[1] http://travisdazell.blogspot.com/2012/11/many-time-pad-attac...
http://www.apprendre-en-ligne.net/crypto/bibliotheque/feiste...
I thought they'd be more directly published (rather than inferred) about the technical methods of VENONA but even there, there's very little. Like, they haven't even released the raw decrypts, just translations. And a bunch of papers congratulating themselves on how awesome they were.
The two-time pad is like Vigenere with a very long key. Offhand, I'd expect that makes the cryptanalysis much more challenging than for Vigenere with a typical short key.
Key size is less interesting to me than the number of times the key is reused.
Kind of, but not quite; the problem with a 2-time pad is key cancellation: c1 ^ c2 = (p1 ^ k) ^ (p2 ^ k) = p1 ^ p2. The xor[0] of two plaintexts is only challenging to cryptanalyze if high-entropy sensitive information happens to line up in just the right way.
0: Technically sum mod 26 or so - xor is sum mod 2 - but the same tricks apply.
At some point your level of sophistication hits that of Lucifer or DES (iterated, nonlinear, nontrivial differential characteristics) and you can declare it "pretty annoying to break" and start comparing it to modern ciphers.
I never thought to go looking before today but I do wonder what the pre-Lucifer designs looked like.
What you have now is a PRNG with a big seed/state vector.