Unfortunately I don’t know how that situation could’ve occurred.
I’ve not heard of this issue showing up before except in situations where the provider schema has changed and where the resolution is therefore to just use the correct provider version.
Your situation is very strange because as far as I can tell there has never been a version 1 of
google_compute_address schema and so I don’t know why the state would be updated that way. On the plus side, I’ve never seen this problem before now so hopefully it was a one-off anomaly for some specific unlikely sequence of events and thus it’s unlikely to happen again. I know that’s not a satisfying answer.
Your workaround of making Terraform forget and then re-import the resource is a find workaround here: that caused the GCP provider to recreate the state based on the upstream resource and the current schema version
0. If you do see this problem appear again, and you’re able to delay the fix to enable some further debugging, it could be helpful to file an issue about it in the GCP provider repository; the maintainers of the provider may have additional context about changes to provider schema etc that I’m not aware of, and so they might be able help identify a specific cause.