# Expected Diagnosis: running-changefeed-with-stalled-progress

The expected result should converge on the case-specific mechanism below while
keeping conclusions bounded by the collected evidence.

## Expected conclusion

In TiCDC Classic Architecture, a Kafka broken pipe, EOF, or connection reset tore down the Kafka client but the producer was not correctly recreated. The changefeed remained Running while checkpoint stopped, and Pause/Resume recovered it by rebuilding the client. Moving to New Architecture is the durable remedy because Classic is not planned for this fix.

## Required evidence chain

- Correlate the checkpoint stall with Kafka transport errors while noting that the control-plane state remains Running.
- Identify the Classic Architecture producer/client recreation defect.
- Use Pause/Resume recovery as evidence that rebuilding the Kafka client restores progress.
- Recommend migration to New Architecture; do not present repeated Pause/Resume as a durable fix.