Kafka `ListOffsets` requests fail when bootstrap broker is not the partition leader

[mdreizin]

What happened?

TablePro successfully connects to a multi-broker Kafka cluster and can discover topics and brokers, but cannot retrieve topic data.

Both of the following commands fail:

DESCRIBE TOPIC "xxx"

CONSUME "xxx" FROM NEWEST LIMIT 100

The failure occurs with:

The broker rejected the ListOffsets request: this broker no longer leads the partition

Cluster metadata is retrieved successfully and shows multiple brokers:

broker 0: <broker-0>:<port>
broker 1: <broker-1>:<port>
broker 2: <broker-2>:<port>

All three advertised broker endpoints are reachable.

The issue appears to be related to ListOffsets requests being sent through the control/bootstrap connection rather than to the current leader of the requested partition.

Steps to reproduce

1. Connect TablePro to a multi-broker Kafka cluster.
2. Verify that the connection succeeds and topics are visible.
3. Run:

DESCRIBE TOPIC "xxx"

4. Run:

CONSUME "xxx" FROM NEWEST LIMIT 100

5. Both commands fail with the ListOffsets error.

Expected behavior

DESCRIBE TOPIC and CONSUME should successfully retrieve the topic metadata and messages.

ListOffsets requests should be routed to the current leader of each partition. If leadership changes, TablePro should refresh metadata and retry the request against the new leader.

Database type

N/A

TablePro version

0.75.0 (131)

macOS version & chip

27.0 / M4 Pro

Screenshots / Logs

The broker rejected the ListOffsets request: this broker no longer leads the partition