How should exposure durations be scaled from studies in mice and rats to humans?

Organogenesis in mouse lasts for about ten days (day 5-15), whereas in humans it lasts for several months (week 8-birth), making it clear that 10 day-exposure in mice is not equivalent to that in humans. There is already a one-day difference between the time of blastocyst formation (Mole et al., 2020; https://doi.org/10.1016/bs.ctdb.2019.10.002). Development of organs and systems differ between humans and rodents (e.g. heart (Krishnan et al., 2014; https://doi.org/10.1038/pr.2014.128), brain (Semple et al., 2013; https://doi.org/10.1016/j.pneurobio.2013.04.001) and the reproductive tract (Cunha et al., 2019; https://doi.org/10.1016/j.diff.2019.07.004)) and the developmental studies should examine and cover both the periods of preimplantation and organogenesis. Due to significant differences between rodents and humans in i.a. toxicokinetics, a non-rodent model is often either required or at least recommended for regulatory purposes.

Understanding developmental sensitivity to preventable environmental threats has progressed significantly. With better understanding comes a wider perspective, and developmental toxicity now includes latent effects that might cause delayed detrimental consequences in adults or in old age, as well as additional effects that are passed down through generations. Although epidemiology and toxicology are increasingly investigating the negative effects of developmental exposures in humans, little progress in protection has been made, and few environmental chemicals are currently regulated to protect against developmental toxicity, whether it is neurotoxicity, endocrine disruption, or other negative outcomes.

My answer #1 is separately and #2 No. The comparison is not simple. This has been a matter of questioning in the toxicology field for many decades. Rodents and human physiology have things in common but also many differences. I think we should primarily consider all the elements separately and the target organs. CDC has a good literature for different toxic components. In animal models we can establish the LC50 but not in humans. Unless the exposure is accidental or measured in certain areas, human exposures also varies by gender and age. In the case of heavy metals, we can estimate the concentrations in water, plants, animals and humans (urine, hair, blood, teeth, nail). Then we can make lab test and evaluate the overall health conditions. Ultimately, collect biopsies and perform pathology studies.   

It is also important to consider the half-life of the toxicant in mice vs humans. For example, dioxin has a shorter half-life in mice than humans because mice have more CYP450 enzymes than humans do. Therefore, some animal models may require a higher initial dose to reflect human exposure--especially in studies where animals are being exposed in utero and through lactation.

There is a good paper summarizing the factors that need to be considered when comparing life stages of (lab) rats and humans - https://www.scielo.br/pdf/abcd/v25n1/en_11.pdf.
Wrt accurate comparison/matching of estimated human exposure with a hazard benchmark value derived from an animal study one should try to align both for temporal aspects (i.e. duration and frequency of exposure) as closely as possible. In this respect, you may find useful a recent ECETOC publication (https://www.ecetoc.org/publication/tr-137-dnel/) reviewing existing approaches to extrapolating from experimental data to human exposure using (data-informed) assessment factors (AFs).