During these past couple of weeks of testing, I've noticed that the installation experience end-to-end seems to take longer using the stub installer versus using the stand alone installer. For example, if you go to nightly.mozilla.org and compare the installation time between the express and standard installers, it just seems like it takes longer, and it is especially noticeable during the "Downloading" phase in the stub installer. We should compare installation times over lots of tries and see if there's a real difference.

What are you comparing the "Downloading" phase of the stub to in relation to the full installer which doesn't download anything?

FWIW my analysis of the raw HTTP logs from download.mozilla.org after the GA test was that across 143K "sessions" (hit to the download page followed by hit to the first-run page from the same IP) that the stub installer downloads were about 6% faster: Average length of full standard sessions, in minutes 4.66002457582 Average length of stub standard sessions, in minutes 4.40170158567

Sorry, the median figures were accidentally left off: Median length of full standard sessions, in minutes 2.78333333333 Median length of stub standard sessions, in minutes 2.6

Are you seeing the same with Aurora?

There are enough additional things that the stub installer does that even if it were slightly longer which it doesn't appear to be per comment #3 that I don't think we should try to compare the two beyond significant differences. Resolving -> wontfix