My experience with JTAG (and it's been a little while since I've used it) is that the device clock must be stopped before you can read the instrumented flip-flops. This is because they're serialized together, meaning you shift them out of the device one at a time using a special JTAG clock signal. You can use JTAG to essentially get a view of the device's state at the moment you stop the clock, but if you keep the clock running you get a trashed device state as you shift the flop values through the serialized chain. (You can also shift values back in and restart the clock, but this may or may not work depending on how many flops you can access via the JTAG chain.)

In other words, I'm not sure how you can use this to modify running devices without being able to stop the clock first. Is this something that JTAG supports now?