I tried to play with different values of min_length parameter as it is shown in the documentation:
stream = split_sentences(nlp, stream, min_length=length)
The text was splitted only in case of min_length=False. In other cases split_sentences does not split the sentences. I tried even min_length=1, but no luck.
Thanks for the report and sorry about that – will look into this! It could either be a side-effect of the recent changes to the split_sentences helper, or the value is overwritten by the split_sents_threshold config setting where it shouldn’t be.
Ok, just tested it and I think what happened is that min_length was accidentally acting like max_length 
Already fixed this for the next release!