A larger key size will mean that you'd have to look at more than just one byte to begin saying anything about what the data could be. This means that even though there are only 128 (normally far far lower for a password) values that each byte COULD be, there would be something like 2keysize-7 possible representations of each of the possible values. This means that an attack like the one described in the blog article would be significantly more difficult.
i don't think so. the attack is extracting a single byte. so without any random iv you will get one of 256 values spread across whatever the block size is. that size doesn't matter - you still have just 256 values to test for. your argument only applies if the rest of the data are appended to the plaintext. [disclaimer: i'm no cryptographer, just using simple logic here]
