I am not sure of the internal details of the delay function, but in general it lets LV do other things until the delay time has passed. The smallest delay is zero, not 1 ms. If you put a zero into the wait function, LV's internal scheduler will check after each iteration of the loop to see if some other node is ready to run. If it is, the other gets a chance to execute, if not the loop processes again immediately. You have no control over the rate, but the zero wait causes the loop to "play nice" with the other parts of the program. It is generally a good practice to have some kind of wait in each loop that is to run in parallel with other parts of the program.
It sounds as though your program may be a good candidate for multiple independent loops. Have one which only handles the TCP communication with the DAQ device. Have another to do the processing and perhaps a thrid for user interface or saving to disk, etc. Each loop can run at different rates, depending on the prioity of the tasks. Pass data and commands between the loops with queues. Look at the state machine examples.
Lynn
