Hello, I'm hitting a persistent CKQuerySubscription save failure on the public database in my Production CloudKit container. I want to flag this is NOT the iOS 26.4 silent-push regression resolved in 26.4.1 (I've confirmed it manifests differently and on devices that predate that bug). App: Win 10 Min (com.luiz.PandaApp) Container: iCloud.com.luiz.PandaApp Current shipping version: 3.2.1 build 420 (approved 2026-05-19) Public database THE ERROR (captured from CKError.userInfo) CKErrorCode 12 (badRequest) CKInternalErrorDomain code 2006 ("BadSyntax") ServerErrorDescription: "attempting to create a subscription in a production container" Execution time per failed attempt: 45-90ms (gateway-level rejection, not data validation). CODE let subscription = CKQuerySubscription( recordType: "GuardianRequest", predicate: NSPredicate(format: "guardianRecordName == %@", me), subscriptionID: "guardian-incoming-(me)", options: [.firesOnRecordCreation] ) let info = CKSubscription.NotificationInfo() info.shouldSendContentAvailable = true subscription.notificationInfo = info try await container.publicCloudDatabase.save(subscription) The .save() throws every time. REPRODUCES ACROSS iPhone on iOS 26.4.2 (i.e., has the 26.4.1 CloudKit regression fix) iPhone on iOS 18.6.x (predates the 26.4 regression entirely) Multiple distinct Apple IDs / userId hashes Fresh installs via TestFlight Updated installs from production CONSISTENTLY: USER_ERROR / BAD_REQUEST. Same error every time. WHY THIS ISN'T THE iOS 26.4 REGRESSION (Thread 820562) That regression caused silent notifications to be DROPPED client-side AFTER successful subscription creation. Our failure is SubscriptionCreate itself being rejected by the CloudKit server before any subscription exists. Different layer, different symptom. iOS 18.6.x devices also fail (they never had the 26.4 token-validation regression code). CLOUDKIT CONSOLE LOGS (sample request UUIDs Apple engineers can lookup) EB30E9D0-8C97-4685-BD16-3E79B3328D6F (2026-05-17 8:08 UTC) EEAA0606-5B50-4640-A950-EEA33B790B8E (2026-05-17, captured in-app) EB4C90CE-9172-4942-B00F-DE4A28388FBA (2026-05-19 7:18 UTC, iOS 18.6.x) F2DF69D9-2339-437A-B59B-7FBD4D57442F (2026-05-19 7:51 UTC, iOS 26.4.x) CloudKit Console → Production → Subscriptions: 0 entries despite 70+ attempts since 2026-05-11. WHAT I'VE VERIFIED / TRIED Schema: GuardianRequest.guardianRecordName and GuardianResponse.requesterRecordName both have QUERYABLE indexes in the Production schema (confirmed in Console). Field types: all STRING. aps-environment entitlement: "production" (changed from "development" in build 420; no change in error pattern). Push Notifications capability: enabled on App ID in Developer Portal AND in Xcode Signing & Capabilities for the main app target. UIBackgroundModes: includes "remote-notification". Re-deployed Production schema from Development on 2026-05-19 (Console reported "no changes"; failure pattern unchanged). Tried with and without info.desiredKeys = []. RecordQuery operations on the same fields and record types in the same container succeed consistently. QUESTIONS Is there a known infrastructure issue affecting public database CKQuerySubscription creation that is distinct from the iOS 26.4 regression (thread 820562)? Is there a per-container configuration that may be in an unexpected state, and if so what should I check? Are there workarounds beyond schema redeployment that have helped other developers in similar situations? Thanks for any guidance.