[beekld]

A synchronizer that fails in a way that will not resolve on retry, such as an HTTP 401, is now parked for a bounded time instead of being retired for the life of the client.

• Synchronizers report Interrupted with UnexpectedFailure set instead of Off. The data system parks the synchronizer, waits a jittered failover delay of 0.5-1s, and starts the next eligible synchronizer after it, wrapping around the list.
• The park delay starts at 5 minutes and doubles up to 1 hour, with jitter. The data system keeps the park count for each synchronizer, so it survives a new instance on each start. A synchronizer clears it with ResetExtendedBackoff when it meets its reset threshold: 60s of healthy streaming, or 2 consecutive successful polls.
• When every synchronizer is parked, the status is Interrupted and the data system waits for the earliest unpark time. Fallback and recovery conditions select from the eligible synchronizers only.
• A synchronizer that fails to build is parked like any other unexpected failure.
• A synchronizer raises the readiness signal only when it delivers a full payload.
• An SDK key change from LDRelayApplySDKKey unparks every synchronizer.

Sorry about the churn in naming ("blocked" -> "parked", "available" -> "eligible", and so on). This work went on in parallel with #466, and while resolving the conflicts I went with the names in the spec.

Note: This does not implement Retry-After or FDv1 Fallback TTL, which are new features that will be added later.

---

Note

Overview
FDv2 no longer permanently retires synchronizers after errors like HTTP 401. They now report Interrupted with UnexpectedFailure, and the data system parks them with escalating backoff (5 minutes, doubling to 1 hour, with jitter) instead of removing them from rotation. When every synchronizer is parked, status stays Interrupted and the loop waits for the earliest unpark time; Off is reserved for explicit Stop().

Synchronizers can clear park escalation via ResetThresholdMet: polling after two consecutive successful polls, streaming after ~60s of healthy operation (including a quiet stream). The FDv2 loop applies a jittered failover delay after parking, treats recovery before fallback when both conditions match, and signals client readiness only for a full payload (IntentTransferFull). LDRelayApplySDKKey unparks all synchronizers so the Relay can retry rejected keys.

Polling and streaming processors, file data source, and DataSynchronizerResult were updated to emit the new flags and default errors to Interrupted rather than Off.

^{Reviewed by Cursor Bugbot for commit e912c37. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 7ac6c1b. Configure here.}