From my experience, I heavily don't recommend blobless/treeless clones for local dev. They should only be used in CI for throwaway build & forget scenario.
When you have blobless/treeless clone on local, it will fetch missing blobs on-demand when doing random `git` operations, in the least expected moments, and it does this very inefficiently, one-by-one, which is super slow. (I also didn't find a way to go back from blobless/treeless clone to a normal clone, i.e. to force-fetch all missing blobs efficiently).
It's especially tricky when those `git` operations happen in background e.g. from a GUI (IDE extension etc.) and you don't have any feedback what is happening.
It might be fine for small repos, but for massive repos, blobless/treeless clones become unmanageable because many git operations become very slow. I added some further links.
From my side, when you have non-trivial sized repo, on local machine one should either use either a shallow, or a full clone (or, if possible, a sparse checkout, but this requires repo to be compatible).
Given that remote git repos are fundamentally incremental in nature, the limit of impracticality is infinite and eventually must equal impossibility.
There is a Moore’s law analogue buried there somewhere in how fast repos grow in relation to network and computing resources (an increase in which of course also makes repos grow faster)
