Well, it's been weeks not a few days and it's not getting any better. Worse if anything,
I have every project checked on the config page, as well as the option at the bottom. WCG is the only BOINC project I run. In fact I only ever have one active project on a machine at any given time. WCG has 100% of this machine's capacity, or it would if it'd use it. btw this isn't some results-rocket trying to punch more than 50 per day. It's an XP 3000+ with 1Gb of ram running Ubuntu linux. It does ok. I checked through the client files and the server seems to think it's only turned on for a couple hours a day at most.
My "on_frac" was about 0.000728 or something when I looked in the "request" file. The "reply" said the same thing.

It's only this AMD machine that's behaving like this, I have two Intel machines running on the same profile, one windohs and one linux, and they both work fine. One of them even gets a error about every 30 or so returned units (thanks to my daughter's internet habits).

<time_stats>
<on_frac>0.000728</on_frac>
<connected_frac>-1.000000</connected_frac>
<active_frac>0.999329</active_frac>
<cpu_efficiency>0.926365</cpu_efficiency>

<project>
<master_url>http://www.worldcommunitygrid.org/</master_url>
<project_name>World Community Grid</project_name>
<symstore></symstore>
<user_name>Dark Angel</user_name>
<team_name>BitBenderTech</team_name>
<host_venue></host_venue>
<email_hash>60b1aec20d901474176fa8927e7e7b47</email_hash>
<cross_project_id>2dfd8ddd2859406a377c18baa360ce62</cross_project_id>
<user_total_credit>149999.365823</user_total_credit>
<user_expavg_credit>468.220463</user_expavg_credit>
<user_create_time>1131744121.000000</user_create_time>
<rpc_seqno>116</rpc_seqno>
<hostid>5027</hostid>
<host_total_credit>52411.882891</host_total_credit>
<host_expavg_credit>185.184243</host_expavg_credit>
<host_create_time>1133205515.000000</host_create_time>
<nrpc_failures>0</nrpc_failures>
<master_fetch_failures>0</master_fetch_failures>
<min_rpc_time>1181542304.399569</min_rpc_time>
<next_rpc_time>1181801201.399569</next_rpc_time>
<short_term_debt>0.000000</short_term_debt>
<long_term_debt>0.000000</long_term_debt>
<resource_share>100.000000</resource_share>
<duration_correction_factor>1.624136</duration_correction_factor>
<sched_rpc_pending>0</sched_rpc_pending>
<verify_files_on_app_start/>
<ams_resource_share>0.000000</ams_resource_share>

I detatched this client from the project and reattached already and that didn't make any difference. I'll try resetting the project and see what that does. At worst I'll only frag one unit anyway so I'll give it a shot. I'll try changing that value back to 1.000000 after the reset if it doesn't do it on it's own.

okay.... your project DCF looks a bit high. If doing 24/7, it confirms you had an event or your benchmark suggests a better performance than actually achieved.

BTW, the last line of your first post: The work quota only gets muscled if you return bad work frequently. It goes up fairly rapidly for each good WU. Somewhere I think knreed posted the gauging table (a wee while back).

Ok then, since there's something else askew I'll stop the client, delete the files and start up again and see what happens.