We use s3 api calls to server read replica and backup/restore pages. When the endpoints fail, which happens when the server is freshly provisioned, these pages cause 500. This commit fixes that by simply swallowing some of the error messages.