Schema Registry 500 errors not reflected in Changefeed status

[ginkgoch]

What did you do?

Create a changefeed on cloud premium with avro or debezium-avro.

What did you expect to see?

• Fail-Fast in Schema Registry: Schema Registry should not hang indefinitely when the backend Kafka cluster is unavailable. It should fail-fast, return a clear error code, and accurately reflect its unhealthy status.
• Accurate Changefeed Status: The changefeed system should detect downstream write failures (500 errors) and immediately update its health status to "Unhealthy".

What did you see instead?

Phenomenon (Symptom)

• Partial Schema Registry Functionality: The Schema Registry RESTful API successfully services read operations (e.g., listing subjects), but any write operations (e.g., registering new schemas) hang indefinitely.
• Misleading Changefeed Status: Downstream changefeeds encounter 500 Internal Server Errors during write attempts, yet the high-level changefeed system status falsely reports as "Healthy" or "Fine."

Root Cause

• Backend Dependency Failure: With the Kafka broker down, Schema Registry cannot write data to its backend _schemas topic (causing write requests to hang), though it can still serve read requests from its local cache.

Versions of the cluster

Upstream TiDB cluster version (execute SELECT tidb_version(); in a MySQL client):

v8.5.7

Upstream TiKV version (execute tikv-server --version):

(paste TiKV version here)

TiCDC version (execute cdc version):

release-version=v26.3.2-8-g9b337c7