You said how you configured the task and the buffer, but you did not say how many points out of the buffer you are currently reading for each iteration. Normally, that error will show up when you are sampling faster than you are reading. You see how if you take 200 samples per second, and read out less than that rate, eventually you will end up with a full buffer. If that happens, you lose a piece of data (it get replaced by new data). Then, when your read VI goes looking for it, it spits out that error.
The trick here will be to read them faster or sample them slower. Is there a reason why they need to come out one at a time? Is there any significance to the 5 ms loop rate (other than it is exactly the period of a 200hz frequency)? You have a bunch of options, your application will determine which path you take.
Best of luck,
~milq