Had a similar experience and had to find this out the hard way. We enabled trace level on a service and forgot to remove for a few days.

Note. Even though the requests were being gziped, that does change your ingested bytes costs.

Two suggestions:

• Add alerts in Cost Explorer just in case something like this happens again.
• Move some trace/debug logs to s3 (batched) and have a short-lived lifecycle policy with a combination of OneZone IA and Glacier Instant Retrieval to save on costs.