What is hard for me to understand is how this could poison spam filters if the volume of it isn't a significant percentage of spam. To keep things from getting too slow, spam filters tend to be pretty primitive, looking at single words or short strings and calculating a probability of an email with all those words being valid or spam. One occurrence of
[email protected] will have a high spam rating, since it would almost never occur in good email. But you could get a lot of spam with a line in the body, "Men the larger wish it were so" before that would raise the likelihood of spamminess, since the filter is probably looking at the single words "men," "larger," and "wish," which are more common in spam, especially when all three are present in the same email, but which certainly occur in valid emails, too.
Sending out spam with innocuous and otherwise uncommon words would tend to lead to good emails being listed as spam. But with the high volume of spam being analyzed, sending spam with innocent words would likely have little effect on a spam filter's ability to recognize true positive spam.
There is always the possibility it is used for list washing -- given the number of English words, a nonsense sentence could uniquely identify an email address if someone reported the spam (or posted it on the internet).