According to our graph[1], between changesets 1cbed63f3eea and 401c71c3732d, we saw a marked drop in TART times for the UX branch, while the m-c branch seemed to stay the same. We should find out why that happened. [1]: http://graphs.mattn.ca/graph-local.html#tests=[[293,1,11],[293,59,11]]&sel=1365611272058.928,1379257972616.8945&displayrange=365&datatype=running Here are the changesets in that range: http://hg.mozilla.org/projects/ux/pushloghtml?fromchange=1cbed63f3eea&tochange=401c71c3732d

Note that this improvement can be seen on the graph server graph happening in the points from May 2nd to May 3rd.

Avi and I suspect that either bug 850918 or bug 826689 caused this performance improvement. I've pushed the May 3rd changeset to try, as well as a push with both of those bugs backed out, and a push with just bug 850918 backed out. Baseline: https://tbpl.mozilla.org/?tree=Try&rev=b2a27f121274 Back out both bug 850918 and bug 826689: https://tbpl.mozilla.org/?tree=Try&rev=65acd3671d73 Back out bug 850918: https://tbpl.mozilla.org/?tree=Try&rev=56cc46ce7b17 Hopefully that gives us some details on what happened here.

I pushed the May 2nd changeset to try as well, for a more complete picture: https://tbpl.mozilla.org/?tree=Try&rev=1187b7b74613

We're not seeing a whole lot of improvement between our May 2nd and May 3rd try pushes... the backfilled data for those dates is a little suspect, so we're retriggering them to hopefully get a clearer sense of the picture.

Matt's awesome, came out of vacation, and helped clean up the data. Some data from some earlier runs were in there, and he's removed them.