Before I get started:
We are a small size office (7 keys) running Windows XP and ArchiCAD10 (and previously 9). We have Dell Dimensions w/ 2GB of RAM w/ gigabit network cards. We have a Gigabit switch and our server has 2GB of Ram.
Our office works almost exclusively in teamwork files and when the whole office is working on one teamwork file it literally takes two to three tries for anyone to do a send/receive operation. Being only human, we have grown used to this and tried to utilize .plc's as much as possible, but at least once a day we end up needing to change workspaces to give up layers/ objects between team members. This past weekend we finally got fed up when 5 hours on a Saturday were spent trying to sign-in and sign-out and we were up against a deadline.
We are in a unique troubleshooting condition as we are an offshoot of a previous firm that also used AC. That firm (which had a different but similar build server, different switch, different cables, different but similar build workstations) had the exact same problems with send/ receive and sign-in/ sign-out. Wanting to know what the problem is once and for all we teamworked a file locally on a harddrive and replicated the multiple user sign-in with weird marques, sporadic layer and layout combos that normally crash on the server. The result was that the C: drive local teamwork file could take anything we dished at it, but the instant we tried it on the server-bust!, SNAFU!! We can only speculate that 1) AC is not coded 100% to work seamlessly on XP based PC's; or 2) There is a problem with the compatability of Dell components with AC; or 3) The network access protocols of teamworking in AC are not up to snuf.
Don't know if anyone else is out there in the same boat, but if you are we have found that some things increase reliability during send/receive.
Use of a marque, even if it surrounds the whole building helps. Being on the native "project 0" level, having only the plan window open, and having minimal layers displaying might help. Expletives have extremely limited results and are not recommended.
FYI we have to save out a pln and then reteamwork it into a plp (with a new name) about once every two weeks when things get really bad.