Ok, so this an accounting issue but the fact that you can charge time elsewhere does not mean very much from POV of a yum _user_. One starts yum and waits for it to finish. That detail that really some subprocess is busy instead does not make an overall wait shorter. If yum itself is responsible only for a small fraction of that time then optimizing it has a negligible overall on its practical use.
So I should go see if I can make the transaction test faster in rpm, too.
Well I'm glad that in addition to working on my own program I have to go hunt down any issues I have in every other library, too.
Sounds like fun. wait wait no, no it doesn't.
:)
-sv