nit: as we would still like to properly validate the Status Code, it might better here to use a "snapshot skip" flag.
Usually, we use the transformers for things we know will change between runs, like randomly generated ID, or things that by design are going to be different than AWS, like endpoints, host values, etc.

When things are not in parity with AWS, we tend to use a pytest marker to indicate that something is not like AWS, and we need to fix it later. It uses JSONPath to indicate which parts of the snapshot to skip.

In this case, you could use the marker like so:

    @markers.snapshot.skip_snapshot_verify(
        paths=[
            # PutBucketMetricsConfiguration should return 204, but we return 200
            "$.put_bucket_metrics_configuration.ResponseMetadata.HTTPStatusCode",
        ]
    )
    @markers.aws.validated
    def test_put_bucket_metrics_configuration(self, s3_bucket, aws_client, snapshot):

You can specify the "snapshot key" as the first part of the path, the one you specify as the first argument of snapshot.match, then the fields you want to skip.
You can add this snapshot skip to the tests having the issue.

You can then remove the transformer, and regenerate the snapshots 😄

Something interesting is happening, and I'm not able to root cause it, at least at this time of the night. I tried to add the skip_snapshot_verify decorator, as you suggested, but my implementation does not work as expected, both in test_put_bucket_metrics_configuration and in test_overwrite_bucket_metrics_configuration tests.

For example:

    @markers.aws.validated
    @markers.snapshot.skip_snapshot_verify(
        paths=[
            # PutBucketMetricsConfiguration should return 204, but we return 200
            "$.put_bucket_metrics_configuration.ResponseMetadata.HTTPStatusCode",
        ]
    )
    def test_put_bucket_metrics_configuration(self, s3_bucket, aws_client, snapshot):
        snapshot.add_transformer([snapshot.transform.key_value("Id")])

        metric_id = short_uid()
        metrics_config = {"Id": metric_id, "Filter": {"Prefix": "logs/"}}

        put_result = aws_client.s3.put_bucket_metrics_configuration(
            Bucket=s3_bucket, Id=metric_id, MetricsConfiguration=metrics_config
        )
        snapshot.match("put_bucket_metrics_configuration", put_result)

        get_result = aws_client.s3.get_bucket_metrics_configuration(Bucket=s3_bucket, Id=metric_id)
        snapshot.match("get_bucket_metrics_configuration", get_result)

After updating again the snapshot, this test runs successfully locally, but fails when run against AWS with the following error:

_____TestS3MetricsConfiguration.test_put_bucket_metrics_configuration _____
>> match key: put_bucket_metrics_configuration
	(~) /ResponseMetadata/HTTPStatusCode 200 → 204 ... (expected → actual)

	Ignore list (please keep in mind list indices might not work and should be replaced):
	["$..ResponseMetadata.HTTPStatusCode"]

I triple checked the path, which seems correct, and also tried some different ones. The test seems to work if I use the same path while adding a transformer with jsonpath, through snapshot.transform.jsonpath. I compared the logic with other usages of the skipper method in the codebase, but I was not able to find significant differences. I searched online, without success.

Are you able to spot any mistake?

This is how the updated snapshot looks like:

  "tests/aws/services/s3/test_s3_api.py::TestS3MetricsConfiguration::test_put_bucket_metrics_configuration": {
    "recorded-date": "13-06-2025, 00:01:27",
    "recorded-content": {
      "put_bucket_metrics_configuration": {
        "ResponseMetadata": {
          "HTTPHeaders": {},
          "HTTPStatusCode": 200
        }
      },
      "get_bucket_metrics_configuration": {
        "MetricsConfiguration": {
          "Filter": {
            "Prefix": "logs/"
          },
          "Id": "<id:1>"
        },
        "ResponseMetadata": {
          "HTTPHeaders": {},
          "HTTPStatusCode": 200
        }
      }
    }
  },

This looks like you updated the snapshot against LocalStack, so that when you run it again against AWS, the snapshot verification fails because it expects the (wrong) status code LocalStack is returning. The "snapshot skip" only works against LocalStack, which is expected.

When you run the test against LocalStack, make sure you're not using the --snapshot-update flag or the SNAPSHOT_UPDATE=1 environment variable. Sorry for the inconvenience!

Totally right, this is now fixed. I expected that AWS tests would have skipped it as well, but it totally makes sense.

Thank you for your patience!