OK, the Librarian stuff seems easy enough to explain: lib/canonical/launchpad/webapp/tales.py does a lot of "if context.icon is not None:" (for icon, mugshot, logo, and probably more). Which of course attempts to load context.icon into memory, and that isn't really necessary.
It may have a desirable "prefetching" effect. OTOH in cases where caches and proxies don't take care of the icon, it's better to incur a bit of back-to-back latency on filling in each icon than to keep the entire page from going to the client until all these latencies have already been incurred.
OK, the Librarian stuff seems easy enough to explain: lib/canonical/ launchpad/ webapp/ tales.py does a lot of "if context.icon is not None:" (for icon, mugshot, logo, and probably more). Which of course attempts to load context.icon into memory, and that isn't really necessary.
It may have a desirable "prefetching" effect. OTOH in cases where caches and proxies don't take care of the icon, it's better to incur a bit of back-to-back latency on filling in each icon than to keep the entire page from going to the client until all these latencies have already been incurred.
I'll treat that as a separate issue.