DOC-2130 Warn against max_in_flight > 1 with batching processors

Add a NOTE to the max_in_flight reference of every batched output (46
connectors registered via MustRegisterBatchOutput) explaining that
setting max_in_flight > 1 alongside a batching block with processors
risks shipping raw, unprocessed messages to the output if a batching
processor errors at runtime. Per CON-461, the underlying behavior
lives in shared benthos framework code; until the next Connect v5
major can fix it, every affected output gets the same advisory.

Implementation: introduce a shared
definitions.max_in_flight_batched override and \$ref it from the 32
batched outputs that lacked an existing max_in_flight override.
Repoint elasticsearch_v8 and elasticsearch_v9 from the now-removed
elasticsearch_max_in_flight definition to the new shared one. Append
the NOTE to the 12 connectors with bespoke max_in_flight prose
(couchbase, cypher, gcp_bigquery, kafka_franz, mongodb, questdb,
redpanda, redpanda_common, redpanda_migrator, salesforce_sink,
snowflake_streaming, sql_raw) plus ockam_kafka.kafka.max_in_flight
where the field is nested. Cloud Connect docs single-source from
this repo, so the same change covers both sites.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: micheleRP

docs-data/overrides.json

@@ -394,6 +394,8 @@ Size in bytes of the per-connection write buffer.
 
 The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
 
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
+
 *Type*: `int`
 
 *Default*: `64`

modules/components/partials/fields/outputs/arc.adoc

@@ -227,6 +227,8 @@ json_map_columns:
 
 The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
 
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
+
 *Type*: `int`
 
 *Default*: `64`

modules/components/partials/fields/outputs/aws_dynamodb.adoc

@@ -211,7 +211,9 @@ This field supports xref:configuration:interpolation.adoc#bloblang-queries[inter
 
 === `max_in_flight`
 
-The maximum number of parallel message batches to have in flight at any given time.
+The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
+
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
 
 *Type*: `int`
 

modules/components/partials/fields/outputs/aws_kinesis.adoc

@@ -206,6 +206,8 @@ Allows you to specify a custom endpoint for the AWS API.
 
 The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
 
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
+
 *Type*: `int`
 
 *Default*: `64`

modules/components/partials/fields/outputs/aws_kinesis_firehose.adoc

@@ -251,6 +251,8 @@ An optional server-side encryption key.
 
 The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
 
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
+
 *Type*: `int`
 
 *Default*: `64`

modules/components/partials/fields/outputs/aws_s3.adoc

@@ -211,7 +211,9 @@ Allows you to specify a custom endpoint for the AWS API.
 
 === `max_in_flight`
 
-The maximum number of parallel message batches to have in flight at any given time.
+The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
+
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
 
 *Type*: `int`
 

modules/components/partials/fields/outputs/aws_sqs.adoc

@@ -201,6 +201,8 @@ item_id: ${! json("id") }
 
 The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
 
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
+
 *Type*: `int`
 
 *Default*: `64`

modules/components/partials/fields/outputs/azure_cosmosdb.adoc

@@ -114,7 +114,9 @@ processors:
 
 === `max_in_flight`
 
-The maximum number of parallel message batches to have in flight at any given time.
+The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
+
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
 
 *Type*: `int`
 

modules/components/partials/fields/outputs/azure_queue_storage.adoc

@@ -114,7 +114,9 @@ processors:
 
 === `max_in_flight`
 
-The maximum number of parallel message batches to have in flight at any given time.
+The maximum number of messages to have in flight at a given time. Increase this to improve throughput.
+
+NOTE: Set `max_in_flight: 1` when a `batching` block with `processors` is configured on this output. If a batching processor (for example `parquet_encode`, `archive`, or `compress`) errors at runtime, Redpanda Connect can write raw, unprocessed messages to the output instead of the encoded batch, producing corrupt downstream data (for example, a `.parquet` file containing raw JSON). Because the batching processor is single threaded, higher values do not improve performance.
 
 *Type*: `int`