We've been running into issues with configurations not deploying due to time out issues, primarily caused by how long the ESX servers take to create a new virtual switch and then port group. We're aware of the switch/port limitations on each host (we're not close yet) and that the LM/ESX servers have to run queries every time so that they don't duplicate vswitch names but as one looks in the recent tasks through the vSphere client it just seems that LM has a set time to complete a task and if it doesn't happen it reverts its steps. I'd say anywhere between 45 and 90 seconds to create a vswitch and/or port and the deployment (and undeployment - see below) times out. These time outs only happen with the creation of a vswitch or a port group, all other activities (copy, register, power, etc.) though do complete a lot faster.
Interestingly enough undeployments can time out too with the primary symptom being abandoned virtual switches which must be manually removed. This can also happen on a failed deployment where LM doesn't really complete its "undo" steps and leaves a few vswitches orphaned. Changing the time out values on the LM web interface has had no impact. ESX servers are 4.1.0.874690, LM is 4.0.4.1406, vCenter 4.1.0.491557. If anyone has a suggestion it would be greatly appreciated...