The functionality in this code seems ripe for epinowcast-do-one-thing-welling - it provides a generic quantiling tool for use on data.tables, which is a capability I’ve seen replicated in a few epinowcast/epinowcast adjacent places now.
What’s the apetite for bring it over, adapting it / the epinowcast code to match?
I think the current namespace quantile.data.table ought to behave a bit more like a generic quantile method + and then some of the fancier stuff that represents a data.table-like interface would be in a separate quantileDT method.