It's not necessarily unsafe, but it is very difficult to do safely. The design suggested is certainly unsafe, however---there's no way to ensure that the values don't lie about their self-reported type and so using that to trigger a coercion is liable to explore things.
If the compiler provides a couple things:
data TypeRepr
typeReprEq :: TypeRepr -> TypeRepr -> Bool
typeOf :: Typeable a => a -> TypeRepr
such that TypeRepr cannot be generated (e.g. faked) by users, typeOf is guaranteed to be genuine, and (this is the hardest) such a thing doesn't violate parametricity then you can use that interface to write safeCoerce :: (Typeable a, Typeable b) => a -> Maybe b
which is guaranteed to only allow the coercion when `a` and `b` actually happen to be the same type.
No, it wouldn't be nice, it would be unsafe. If one day you end up adding a new type in your container, you need to update your runtime check as well or your code will fail in mysterious ways.
Erasure keeps you honest by asking you to think carefull about the types so that they can be checked by the compiler, and once the compiler has done this verification, you are guaranteed that your code will work.
Any language feature that encourages the use of reflection , such as reification of types, should not be supported by a language that wants to claim to be sound.