BOLD Signal Falling to Zero: What On Earth Is Happening Here?

I'm getting a very odd result: when I plot the ROI BOLD series after denoising, I get a very strange artifact: the BOLD signal will sometimes simply drop to 0, without any variation for seemingly random periods of time. If you imagine a normal, noisy looking waves, with large sections clipped out, you have a sense of what's going on.

This seems to occur in datasets at random: some are extremely vulnerable to it, while others don't show this artifact at all.

All the best
~Thomas