chore: announce kafka connector group ID generation strategy change

docs/components/connectors/out-of-the-box-connectors/kafka.md

@@ -291,6 +291,7 @@ In the **Authentication** section, select the **Authentication type**. If you se
 
 In the **Kafka** section, you can configure the following properties:
 
+- **Consumer Group ID**: Set the consumer group ID for this connector. Always provide an explicit, stable value that identifies the logical consumer group (for example, `my-app-order-processor`). If you leave this field empty, the connector auto-generates an ID from its internal deduplication key. That generated ID can change across connector upgrades, including from 8.8 to 8.9, causing Kafka to treat the connector as a new consumer group and potentially replay already processed messages.
 - **Schema strategy**: Select the schema strategy for your messages.
   - Select **No schema**, **Inline schema** for Avro serialization.
   - Select **Schema registry** If you have a Confluent Schema Registry.
@@ -538,6 +539,10 @@ If any of the field is not populated, you must configure your security method fo
   enable.auto.commit=false
   ```
 
+:::caution
+The `group.id` value above is auto-generated when no explicit **Consumer Group ID** is configured in the connector. This generated ID is derived from the connector's internal deduplication key and can change across connector upgrades, including from 8.8 to 8.9. When the group ID changes, Kafka treats the connector as a new consumer group, which means committed offsets are not reused and messages may be replayed. To avoid this, always set an explicit **Consumer Group ID**. You can [look up existing consumer groups](https://docs.confluent.io/kafka/operations-tools/manage-consumer-groups.html#list-groups-and-view-offsets) to find the current group ID in use.
+:::
+
 ### What is the precedence of client properties loading?
 
 Properties loading consists of three steps:

docs/reference/announcements-release-notes/890/890-announcements.md

@@ -543,6 +543,26 @@ For more information on optimizing connector performance with virtual threads, s
 </div>
 </div>
 
+<div className="release-announcement-row">
+<div className="release-announcement-badge">
+<span className="badge badge--change">Change</span>
+</div>
+<div className="release-announcement-content">
+
+#### Kafka consumer connector: auto-generated consumer group ID format changed
+
+When upgrading from Camunda 8.8 to 8.9, the auto-generated Kafka consumer group ID changes format. This affects Kafka consumer connectors only when no explicit consumer group ID is configured. The change is a side effect of a runtime improvement that enables cross-version connector deduplication.
+
+- **8.8 format:** `kafka-inbound-connector--<processDefinitionKey>-<elementId>` (numeric key)
+- **8.9 format:** `kafka-inbound-connector--<bpmnProcessId>-<hash>` (stable process ID)
+
+Because Kafka treats a changed group ID as a brand-new consumer group, affected connectors do not reuse their committed offsets after the upgrade and may reprocess messages.
+
+**Action required:** Before upgrading to 8.9, set an explicit **Consumer Group ID** in each Kafka consumer connector configuration to preserve the existing consumer group identity. You can [look up existing consumer groups](https://docs.confluent.io/kafka/operations-tools/manage-consumer-groups.html#list-groups-and-view-offsets) to find the current group ID in use. See the [Kafka connector documentation](/components/connectors/out-of-the-box-connectors/kafka.md?kafka=inbound) for details.
+
+</div>
+</div>
+
 ## Data
 
 <div className="release-announcement-row">