we did fix an issue where timeouts were ocurring if a provider ended a stream without a

finish_reason