Minor: Do you think the risk is low that someone misconstrues seed here for reproducibility / something involved in numpyro.sample() vs. interpreting post_seed_latent_infections as the infections after seeded infections $I_{v:0}$ have been incorporated?
I think the risk is low, given that those who might be interested in using this enough to peruse the source code would understand this.
_Originally posted by @AFg6K7h4fhy2 in https://github.com/CDCgov/multisignal-epi-inference/pull/204#discussion_r1647704016_