Comment 3 for bug 662523

Moved from EDGE_SERVICE_ROOT to LPNET_SERVICE_ROOT a few days ago (as edge was completely broken) and it solved the issue for a while.
This morning, at ~7:45am (5:45 UTC), it restarted to timeout:

2010/10/26-04:31:08 66 sec
2010/10/26-04:46:01 59 sec
2010/10/26-05:01:32 90 sec
2010/10/26-05:15:48 46 sec
2010/10/26-05:30:46 44 sec
2010/10/26-05:45:42 41 sec
2010/10/26-06:01:29 88 sec
2010/10/26-06:15:57 55 sec
2010/10/26-06:31:14 73 sec
2010/10/26-06:46:29 87 sec
2010/10/26-07:01:58 116 sec
2010/10/26-07:16:57 116 sec
2010/10/26-07:31:15 74 sec
2010/10/26-07:47:52 170 sec <===
2010/10/26-08:02:52 170 sec
2010/10/26-08:18:03 182 sec
2010/10/26-08:32:57 175 sec
2010/10/26-08:47:59 177 sec
2010/10/26-09:03:02 181 sec
2010/10/26-09:17:57 175 sec
2010/10/26-09:32:59 177 sec
2010/10/26-09:47:52 170 sec

i call getBuildSummariesForSourceIds(source_ids = chunk) with chunks of 25 src ids, it worked for month on edge, it's no longer usable because of timeouts. now it's the same on the production server.
I just lower it down to chunks of 15, but that means more requests and longer sessions.

i've already implemented caching to help solve this: http://paste.ubuntu.com/520087/
i'm out of options.