W2K3 TS and wks losing connection to network share

I have several issues that appear to be network related.

Current scenario,

Two w2k3 TS standard servers (both AD one running exchsrvr2k3). Both use dual port Intel pro MT/1000 with Intel adaptive load balancing team configured on each. Also a w2k member server running pervasive engine (nothing else). Data is currently stored on 14bay storage array. AV software is trendmicro scanmail for exchange, server protect and officescan for wks. Using all Dell switches, (4x)24-100s with 1000 uplink unmanaged, (1x)24-100 with 1000 uplink managed, all connected to 12 port gigabit backplane managed. Wks are varity of XP and 98 systems (around 60).

The problem existed before using Intel load balancing software. Most likely, not Intel driver or software problem.

A. TS2 client connection to TS1 access shares from TS1 (ts profiles, data, exchange) loss connection to resource randomly.
B. XP and 98 Client workstations also lose network connection to server shares. This does not happen as often or more likely is not noticed only because most are using TS rather than their own wks, but it does happen.
C. Netdiag and DCdiag even in verbose mode yield no errors with AD or DNS.
D. Stressing network with multiple concurrent large file transfers do not produce errors in Intel NICs logs nor in Dell switches.
E. Netmon capture also doesn't produce any obvious errors even when a file access error occurs.

Other symptoms like TS profiles not being loaded consistently over network, sometimes it connects to user TS profile share sometimes not. Always works fine on server that shares the profiles. Mapped network drives are available even if profile fails to load. During file use over network from XP client or TS session saves will not succeed, if user waits a bit it sometimes saves. Also opening files sometimes comes up with unrecognized file format, i.e., user opens a pub file and publisher states it is not a recognized file format or network resources does not exist or you don't have access to file.

It almost seems that it is a security issue as no errors appears at the hardware level (switches/NIC) or in event logs. It is like the FSMO running exchange is disconnecting clients and failing to reconnect (with appropriate security). Sec channel problem? Tests did not reveal any problem with this, although they are just tests and are not real world stress.

Anyone else having same problems with network resource availability issues?