How would it go if compression of your blob happened as a background task?

Store some meta-data about the blob somewhere, indicating whether or not
it's been compressed.

Blobs get written in full-size format to begin with. Then, as part of the
sweep, or a cron job, or some other scheduled task, look for blobs that
haven't been compressed, zip them up, and change the meta-data to suit.

Just an idea.

N.