S3 gzipped content automatically unzipped when downloading with a client software

Hi, we have quite a large amount of files stored to S3 and have used it without any problems. Files have Content-Encoding gzip, Content-Type application/octet-stream. What has now changed is that when accessing files using S3 URL with (for example) an HTTP tool most of the files are fine but some are served unzipped, having no Content-Encoding in the header. Data in S3 has not changed, earlier these same files were downloaded fine. This problem seems to happen only when using HTTP, HTTPS is fine. Unfortunately we have an older application using HTTP that now fails. Any ideas what is going on and how to prevent this?