Sure, anyone could write it today, but your proposal is specifically defining a family of APIs and adding it to concrete types, and one of the questions we discussed was "are there any other stdlib types that it makes sense to add this to" (see John's reply as well for some others).
As for not making yourself available to write the implementation, you pretty much did already write it in your reply 
I'm not asking for specific changes here, it's more of a thought exercise. Identity as it's being defined by your proposal can be used to define a different notion of equality of two values, and there's a fairly straightforward generalization of that concept that extends to hashability (since the two would need to be kept in sync)—it's that the hash value would be determined using the same values being compared by isIdentical(to:). If the identity of a value is useful for comparison, then I'm trying to discern whether there are situations where the corresponding "identity hash" would also be useful. The use case "I have a compile-time static value that I want to track by its location in memory instead of its value" is one that comes to mind. Some kind of "intern table" for large CoW values might be another where it's useful to track the identity itself (or a value derived from it) instead of just whether two already-known values are identical.