Sparse split multi + filter changed loci

Hi hail team! I have a basic question about the filter_changed_loci option of haill.experimental.sparse_split_multi. Would setting this option to True ever remove ref block information?

Just looked at the implementation, and it looks like the answer is yes, but probably it’s hard to construct an example that does – as long as any of the alleles from a multiallelic site survive that filter, the ref blocks will be present.