Inconsistency in Content-Disposition format

Dataverse returns Content-Disposition header for a file in two different formats depending on where the file is physically stored. E.g.:

import requests

url = "https://dataverse.harvard.edu/api/access/datafile/3315157"
req = requests.get(url, allow_redirects=True)
print(f"File (id:3315157) - tab version not on S3: {req.headers['Content-Disposition']}")
req = requests.get(url + "?format=original", allow_redirects=True)
print(f"File (id:3315157) - original version on S3: {req.headers['Content-Disposition']}")

yields:

File (id:3315157) - tab version not on S3: attachment; filename="Karnataka_DD%26FS_Data-1.tab"
File (id:3315157) - original version on S3: attachment; filename*=UTF-8''Karnataka_DD%26FS_Data-1.xlsx

Note that in both cases filename has escaped and encoded characters per RFC 5987. However, only the latter format of the header is the correct ("filename*" indicates that name is encoded, see e.g. Content-Disposition def)
Related PRs: #4542 and #7503

Reproduce the two responses with the Python requests example and compare the Content-Disposition headers for the default and original formats. Trace the API response handling for files stored in different locations, then verify that both paths use the RFC 5987 filename* form consistently.