aws-vpc-operator broken after AWSCluster CRD upgrade (subnets[*].resourceID?)

[AndiDog]

Bug description

CAPA private cluster E2E tests fail because aws-vpc-operator throws an error when cleaning up subnets (chat).

AWSCluster.infrastructure.cluster.x-k8s.io "t-b0r9t061akd1lzbl0j" is invalid: spec.network.subnets[1]: Duplicate value: map[string]interface {}{"id":""}

First guess: the breaking change to require subnets[*].resourceID hasn't been applied when bumping to the v1beta2 CRD types.

Designed behavior

Create and delete stuff correctly.

Reproduction Steps

PR comment to only trigger private tests (takes roughly an hour):

/run cluster-test-suites TARGET_SUITES=./providers/capa/private

Technical information

Since the operator doesn't do much on subnets, I'm wondering why the reconciliation code even needs to touch those fields. It's supposed to only patch in route table IDs and stuff, rather than overwrite ID fields. And maybe the whole functionality can be replaced by CAPA tagging the subnets correctly, removing our whole subnets reconciliation code?