Return of XBLR documents is unreliable

GET request to
https://document-api.companieshouse.gov.uk/document/KdaHaSdSU004VC1zokBXaNV5aJ6HyOVPg2vPpQYTTak/content

with Accept:‘application/XBLR’ in headers
SOMETIMES returns XBLR, SOMETIMES returns pdf (more often than not it is pdf).

Are there servers running different versions of the API meaning I’m sometimes hitting a server that does not support XBLR?