Unsure if this would constitute a breaking change or a bug fix? Since technically it's sort of both?
Previously Env set variables to be a required union, regardless whether the variable existed across all environments or not. This has now been changed such that variables in Env are still a union, but are now optional unless that variable exists across all named environments.

mh.... the same can be said to any other binding no?

For example I could have something like:

	"kv_namespaces": [{
		"binding": "MY_KV",
		"id": "xxx"
	}],
	"env": {
		"my_env": {
			"kv_namespaces": []
		}
	}

with the previous iteration of wrangler you'd get MY_KV: KVNamespace while now you'd get MY_KV?: KVNamespace (since MY_KV is not declared for the my_env environment) no? (so here again users could go from a normal variable before to an optional one after)

Yes pretty much. Based on your config above my concern is that users would now have to modify their code to account for these new / fixed types.

For comparison sake, here is npx wrangler@latest types vs pnpm run cf-typegen on this branch with the config you shared above.

Before

After

Thanks for the clarification, yeah it's pretty much as I thought...

I agree that there's the risk of breaking folks... I personally would think that this is ok since this makes the types more accurate (so it could be considered a bugfix?) but you can ask for a second opinion from the team someone might think of this differently (we could make this opt-in if we were really concerned 🤔)