Conflict between the router and HMR

Every time I change route using ``, all my components get _reconstructed_ (not updated, not even mounted again, but the class constructor is called). It makes it impossible to keep component states when changing routes without putting everything in store.

Is that really intended? Shouldn't client-side routing just (un-)mount or update components w.r.t url?

Sorry for asking here, but the downside of having a hand-made router is that there is no help to find online by people using another generator.

Start by reproducing the behavior when changing routes with , then inspect how the router and HMR interact during navigation. Done means determining whether component constructors are incorrectly called on route changes and documenting or fixing the expected mount and state-preservation behavior.

Written by the indexing model from the issue text.