i watched this all the way through (interesting stuff around slide 15) and the one question i have is how do they control sequential writing to a singular 'hays

i watched this all the way through (interesting stuff around slide 15) and the one question i have is how do they control sequential writing to a singular 'haystack' file that is being (artificially) capped at 10GB. they are using offsets to seek in this large file to limit their exposure to unnecessary file metadata.