This reminds me of those daze when they did experiments with spiders.
They dosed them with LSD and took pictures of the webs they wove
afterwards. I've intentionally deleted the end-points. I also sent
this to both Qwest and SPeakEasy over two weeks ago. These routes are
persistent.
It's not their fault (not counting poor decisions of where to peer or buy
transit), it's INAPs.
---Gets to San Francisco on the first hop (looks like an overloaded router)
3 gw-081-176.sfo1.dsl.speakeasy.net (216.27.176.1) 24.758 ms 118.038 ms 27.967 ms---Plays around on the InterNAP.
4 border3.fe5-3.speakeasy-9.sff.pnap.net (216.52.86.28) 20.485 ms 20.795 ms 21.185 ms
5 core4.fe0-0-bbnet1.sff.pnap.net (216.52.80.6) 22.027 ms 20.601 ms 21.605 ms
Speakeasy uses InterNAP for transit... Though why a DSL provider would
pick a company who's model for optimizing traffic is far more useful
outbound then inbound is beyond me. Maybe INAP gave them some $200/meg
transit to try and stay in business...
---Finally gets to QWest (used to be USworst ... may still be) but still
hasn't left SF
And InterNAP uses QWest for transit...
---Then goes down to Los Angeles, for some sunshine.
Semi reasonable on qwest's part... Can't afford to peer with everyone
everywhere can you... 
---Where it finally crosses over to AboveNet
12 sjc2-lax1-oc12.sjc2.above.net (208.184.102.145) 51.917 ms 45.190 ms 47.872 ms
13 sjc2-lax1-oc48.sjc2.above.net (208.185.156.125) 49.168 ms 52.816 ms 62.538 ms
Well you did find some broken DNS at AboveNet... (or else they're trying
out some new IP over teleportation technology)... From tracerouting around
a little bit it looks like that hop 13 is actually an OC48 between core1
and core4 in sjc2.above.net.
---and heads back to San Jose
14 core5-sjc2-oc48.sjc1.above.net (216.200.0.177) 48.298 ms 80.466 ms 43.326 ms
15 main2-core5-oc12.sjc1.above.net (209.133.31.190) 44.550 ms 64.522 ms
47.930 ms
Over to another facility in San Jose...
Coming back seems to be much more intelligent (less drifty);
3 main3-main4-fe.sjc1.above.net (209.249.192.4) 7.212 ms 5.777 ms 7.242 ms
4 core1-main1-oc12.sjc1.above.net (208.185.175.245) 7.237 ms 6.944 ms 7.232 ms
5 sl-gw8-sj-0-1.sprintlink.net (144.232.192.129) 7.231 ms 6.973 ms 7.233 ms
6 sl-bb21-sj-13-0.sprintlink.net (144.232.3.113) 7.230 ms 6.973 ms 7.245 ms
7 sl-gw12-sj-10-0.sprintlink.net (144.232.3.142) 7.227 ms 6.926 ms 7.241 ms
8 sl-internap-40-0-0.sprintlink.net (144.232.217.18) 18.254 ms 17.886 ms 18.251 ms
9 border4.ge2-0-bbnet1.sff.pnap.net (216.52.80.9) 18.170 ms 17.905 ms 21.890 ms
10 spk-1-sfo.dsl-isp.net (216.52.86.1) 29.110 ms 21.618 ms 32.778 ms
11 dsl027-176-139.sfo1.dsl.speakeasy.net (216.27.176.139) 40.001 ms 39.886 ms 40.082 ms
Well who's to say that InterNAP didn't find a better path... It's always
possible that using Qwest and peering with AboveNet in LA is better then
using Sprint and peering with AboveNet in the bay area...
That being said, it's more likely then not just minus points for the
ASimilator... Can't win 'em all...