As there are only two of each character in the set which is shuffled ((@chars) x 2), there can of course also only occur two in a row at maximum.

OTOH, this technique doesn't allow more than two of any character in the result (like AACBAFFDCAEC), so it might introduce an unwanted bias... (you didn't explicitly say whether this is desired or not, and it cannot be inferred from your valid samples).