We extend the generic framework of reproducibility for reuse of randomness in multi-recipient encryption schemes as proposed by Bellare et al. (PKC 2003). A new notion of weak reproducibility captures not only encryption schemes which are (fully) reproducible under the criteria given in the previous work, but also a class of efficient schemes which can only be used in the single message setting. In particular, we are able to capture the single message schemes suggested by Kurosawa (PKC 2002), which are more efficient than the direct adaptation of the multiple message schemes studied by Bellare et al. Our study of randomness reuse in key encapsulation mechanisms provides an additional argument for the relevance of these results: by taking advantage of our weak reproducibility notion, we are able to generalise and improve multi-recipient KEM constructions found in literature. We also propose an efficient multi-recipient KEM provably secure in the standard model and conclude the paper by proposing a notion of direct reproducibility which enables tighter security reductions.