[perfectra1n]

Fixes a batch of six open issues, one commit per issue. Every fix ships with a regression test that fails without it; the whole branch is green on cargo test/clippy -D warnings/fmt/gen-check and the cognitive-complexity ratchet.

Fixes

• fix(api) RepositoryReplication with different source and destination gets rejected by admission hook #248 — replication rejected when source ≠ destination. backend_target_key keyed S3 on bucket+prefix only (dropping endpoint+region), Azure dropped storage_account, Filesystem dropped volume — so two distinct backends sharing a bucket/container/path name collided and the webhook rejected a valid RepositoryReplication between different S3 providers. Now folds every storage-identifying field into the key.
• fix(controller) SnapshotPolicy's defaultDeletionPolicy doesn't work #238 — SnapshotPolicy.defaultDeletionPolicy was write-only. create_scheduled_backup hardcoded deletionPolicy: None, so the mutating webhook always stamped its origin default (Delete). Now resolves the policy and threads its default onto the child Snapshot before admission.
• fix(controller) ClusterRepository keeps stale suspended conditions after resume #245 — stale Suspended conditions after resume. io::set_ready was called only in the suspend branch, so a healthy resume left Ready/Reconciling/Stalled frozen at the suspend generation (Flux wait: true hangs), and a never-suspended repo never got them. New io::ready_outcome_for_phase is layered onto every terminal status write in both repo reconcilers.
• feat(metrics) Backups Ok metric showing a number greater than number of policies #196 — "Backups OK" exceeded the policy count. The panel summed a durable per-run counter over 24h. Adds kopiur_snapshotpolicy_last_backup_success (0/1 per policy) and retargets the panel to count(... == 1) → "Policies Healthy".
• fix(mover) Bootstrap result exceeds the ConfigMap 1MiB limit at ~500+ snapshots, permanently wedging the repository at Bootstrapped: False #237 — bootstrap ConfigMap > 1MiB permanently wedged the repo. MAX_RETURNED_SNAPSHOTS caps entry count, not size; ~500 full entries exceeded etcd's 1MiB limit, the write 422'd, the Job still exited 0, and the controller looped ResultPending forever at Bootstrapped: False. Strips the three controller-unused heavy fields (~400 B/entry) + a size-budget backstop + a loud write-failure log.
• fix(mover) SnapshotPolicy verification.quick does not target the snapshot with --scope #250 — quick verify wasn't scoped to the policy. kopia snapshot verify ran with no --sources, so a per-policy quick verify verified the whole shared repository (and verifyFilesPercent sampled every identity's data). Now scopes the quick tier to the run's resolved identity.

Noted follow-ups (out of each issue's core scope)

• SnapshotPolicy's defaultDeletionPolicy doesn't work #238: a manually kubectl create'd Snapshot with a policyRef still won't inherit the default — the Snapshot admission webhook is synchronous/client-less; fixing needs it async+client.
• ClusterRepository keeps stale suspended conditions after resume #245 / SnapshotPolicy's defaultDeletionPolicy doesn't work #238: hermetic-unit-guarded here; e2e suspend→resume and defaultDeletionPolicy end-to-end scenarios are the ideal full guards.
• Bootstrap result exceeds the ConfigMap 1MiB limit at ~500+ snapshots, permanently wedging the repository at Bootstrapped: False #237: controller-side bounded-retry escalation of ResultPending → Warning (the strip + size guard already make the oversized case unreachable, so this is defense-in-depth).

Closes #248, closes #238, closes #245, closes #196, closes #237, closes #250.

[greptile-apps]

Greptile Summary

This PR fixes several backup operator regressions across validation, scheduling, repository status, mover jobs, verification, and observability. The main changes are:

• More precise backend identity keys for replication validation.
• Scheduled Snapshots inherit SnapshotPolicy.spec.defaultDeletionPolicy before admission.
• Repository and ClusterRepository status writes refresh kstatus Ready, Reconciling, and Stalled conditions.
• Quick verification runs are scoped to the policy’s resolved kopia source.
• Bootstrap catalog results are slimmed and size-guarded before writing to the result ConfigMap.
• Policy health dashboards now use a per-policy last-backup-success gauge.

Confidence Score: 5/5

Safe to merge with low risk.

The changes are targeted fixes with tests covering the affected validation, controller, mover, kopia argument, metrics, and dashboard paths. No blocking correctness or security issues were identified. The previous defaultDeletionPolicy fallback issue is addressed by propagating policy read errors instead of creating a Snapshot with None.

No files require special attention.

T-Rex Logs

What T-Rex did

• Executed the cargo test for the changed crates and produced a proof log that records the command, working directory, cargo test output, EXIT_CODE: 0, and ELAPSED_SECONDS: 11.
• Validated that no narrowed retry was needed because the highest-value requested command passed.

_{Ran code and verified through T-Rex}

Important Files Changed

Filename	Overview
crates/api/src/validate/repository.rs	Expands replication backend identity keys to include storage-target discriminators for S3, Azure, filesystem, and SFTP.
crates/controller/src/snapshot_schedule.rs	Resolves SnapshotPolicy.defaultDeletionPolicy before creating scheduled Snapshots and propagates read errors.
crates/controller/src/repository.rs	Adds kstatus condition updates to namespaced repository terminal status paths.
crates/controller/src/cluster_repository.rs	Adds matching kstatus condition updates to cluster repository success, failure, and probe paths.
crates/controller/src/metrics.rs	Adds a per-policy 0/1 gauge for latest terminal backup success.
crates/kopia/src/client/mod.rs	Adds VerifyOptions.sources and emits --sources for scoped kopia snapshot verify calls.
crates/mover/src/bootstrap.rs	Slims bootstrap catalog entries and adds a serialized-result size budget.
crates/mover/src/main.rs	Applies the bootstrap result size guard and scopes quick verification to the resolved identity.
deploy/dashboards/kopiur.json	Updates the policy-health dashboard panel to use the new last-backup-success gauge.

_{Reviews (2): Last reviewed commit: "fix(controller): fail the fire on a poli..." | Re-trigger Greptile}