Currently, if you visit a git-ssb-web repo page you might end up downloading large ssb blobs, if some git object needed to render the page is stored in a large packfile (i.e. was part of a large git push payload). A better behavior would be to only have to download large packfiles if you are actually cloning/fetching the repo, and the web UI could be kept fast. One way to help ensure this would be to split a large packfile so that objects likely to be needed just to render the web UI go in one (small) pack, and the rest of the objects go in another (large) pack. Then loading the web UI would be more likely to only trigger a request for the small packfile. Specifically, objects which should go in the small pack are the blob for the project's readme, the root tree, commit objects, image blobs referenced in the readme, trees and readmes in subdirectories, etc.

This issue is relevant mostly to importing large repositories. However, even for existing git-ssb repos, it might make sense in some cases to split out the large blobs into a separate packfile from the commit and tree objects, so that viewing e.g. a repo's latest commit message doesn't trigger fetching large packfiles.