The number of "lost connection" messages in server.log seems a little excessive, lately.

uniblab:Swim2 steve$ for i in `seq w 1 20`; do echo $i: `cat server.log | grep 2013-02$i | grep lost\ connection | wc -l`; done
01: 268
02: 393
03: 505
04: 228
05: 292
06: 187
07: 248
08: 227
09: 365
10: 451
11: 374
12: 344
13: 406
14: 522
15: 730
16: 1000
17: 730
18: 1821
19: 5015
20: 2937
uniblab:Swim2 steve${noformat}