scaleway/pigeon-cluster v0.2.1
v0.2.0 (ADR-0144) gave each pigeon-cluster job its own `job_policy`, named `"${name_prefix}-${job}-iam-policy"`. That's the exact same formula compute-instance's own internal self-delete `iam_policy` composition already uses for the same job -- both resolve to the identical final Scaleway policy name once iam-policy's own `-policy` suffix is applied. Scaleway rejects the second create with a 409 (`resource policy: resource already exists`), confirmed live against a real `terragrunt apply` of `v0.2.0`.
`job_policy`'s name now gets a distinct `-work-` segment so the two policies -- one scoped to the job's actual work permissions (object storage, etc.), one scoped to just self-delete -- coexist without colliding.
No interface change, no consumer update needed beyond bumping the version pin.