Add known issue for AWS S3 performance regression in 8.12 (#865) (#867)

* Add known issue for S3 performance regression.

* Update docs/en/ingest-management/release-notes/release-notes-8.12.asciidoc

Co-authored-by: David Kilfoyle <[email protected]>

* Change wording slightly.

---------

Co-authored-by: David Kilfoyle <[email protected]>
(cherry picked from commit 6a059ce)

Co-authored-by: Craig MacKenzie <[email protected]>

docs/en/ingest-management/release-notes/release-notes-8.12.asciidoc

@@ -230,6 +230,26 @@ sudo systemctl start elastic-agent
 
 ====
 
+[[known-issue-37754]]
+.Performance regression in AWS S3 inputs using SQS notification
+[%collapsible]
+====
+
+*Details*
+
+In 8.12 the default memory queue flush interval was raised from 1 second to 10 seconds. In many configurations this improves performance because it allows the output to batch more events per round trip, which improves efficiency. However, the SQS input has an extra bottleneck that interacts badly with the new value.
+
+For more details see {beats-issue}37754[#37754].
+
+*Impact* +
+
+If you are using the Elasticsearch output, and your configuration uses a performance preset, switch it to `preset: latency`. If you use no preset or use `preset: custom`, then set `queue.mem.flush.timeout: 1` in your output configuration.
+
+If you are not using the Elasticsearch output, set `queue.mem.flush.timeout: 1` in your output configuration.
+
+To configure the output parameters for a {fleet}-managed agent, see <<es-output-settings-yaml-config>>. For a standalone agent, see <<elastic-agent-output-configuration>>.
+
+====
 
 [[known-issue-sec8366]]
 .{fleet} setup can fail when there are more than one thousand {agent} policies