[beekld]

Summary

Applies the RETRY behavior to FDv2 synchronizer orchestration in the server SDK. A synchronizer that fails is parked and retried later, instead of being blocked for good.

Parking

• A synchronizer that fails unexpectedly is parked: a terminal error, an exception from next(), a failure to build, or shutting down on its own. The park delay starts at 5 minutes, doubles with each failure up to 1 hour, and has jitter. After a park, the next synchronizer starts after a failover delay of 0.5-1 seconds.
• A synchronizer's park count resets when it meets its reset threshold. For polling, that's two successful polls in a row. For streaming, it's staying connected for 60 seconds.
• When every synchronizer is parked, the data system waits for the earliest unpark time. Synchronization stops only when the data source is closed.
• When the fallback condition is met but no other synchronizer is available, the current synchronizer keeps running.

Directed fallback

• The fallback to FDv1 now ends after a TTL, which defaults to 1 hour with jitter. The data system then restarts the FDv2 synchronizers.
• A directive received during the fallback extends the TTL.
• A directive with no FDv1 fallback synchronizer configured no longer halts the data system. It reports INTERRUPTED and restarts the FDv2 synchronizers when the TTL ends.

Conditions removed

• The Conditions class and FDv2DataSourceConditions are removed. They weren't compatible with the new parking and FDv1 fallback timers, so the synchronizer loop is simplified to check all of its timers each time it wakes.

Not included

• Honoring Retry-After.
• Reading the TTL from the response. Every directive uses the 1-hour default.

---

Note

Overview
Replaces the old condition/timer model (FDv2DataSourceConditions, FDv2DataSource.Conditions) with a single synchronizer loop that wakes on results or deadlines (fallback, recovery, unpark, FDv1 TTL). Failed synchronizers are parked with exponential backoff (5 min → 1 h, jitter) instead of being blocked for the process lifetime; terminal errors and build failures park the source rather than permanently blocking it.

Adds AsyncLatch so close() and timed waits can race CompletableFuture.anyOf without leaking completion nodes on long-lived streams.

FDv1 fallback is time-bounded (default 1 h, jitter): SourceManager tracks end time, blocks FDv2 sources during fallback, and returns when TTL expires. Server directives extend TTL or switch to FDv1 without halting the whole data system when no FDv1 synchronizer exists (reports INTERRUPTED and waits for TTL).

Reset threshold is propagated via FDv2SourceResult.resetThresholdMet — polling (2 successes), streaming (60 s healthy), FDv1 adapter (2 VALID) — and resets park counts. HTTP/transport classification (FailureClass) turns unexpected errors into terminal results in polling, streaming, and the FDv1 synchronizer adapter.

close() releases the latch, shuts down sources, and joins the run thread (5 s cap) before reporting OFF.

^{Reviewed by Cursor Bugbot for commit 94f7478. 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 0dafa42. Configure here.}