Let's improve our test coverage as a follow-up to: https://github.com/kubernetes-sigs/controller-runtime/pull/2321 The PR description describes the deadlock we encountered in detail. The following describes the test we need to ensure this doesn't happen again. * Create a CRD with apiVersion v1 * Create CRs for this apiVersion * Update the CRD to introduce apiVersion v2 (which should be also the new storage version) * Setup a manager which: * implements conversion for this CRD * reconciles / watches the CRD * Start the manager and ensure it works without a deadlock