I think that is worth considering. The main advantage is, that it allows new nightly features without thinking and thus enables more experimentation.
Of course that would come at the cost of complicated packaging.
But I like the rust approach of babysitting the programmer.
Rust often has the problem that not enough programmers are testing nightly features because of the complicated toolchain and feature annotations. Those features then get permanently stabilized and someone discovers a problem, but then it’s too late, because 1.0 can almost never be changed. However I consider that a wortwhile tradeoff.