This was quite the learning experience, as I forced my reactor to suffer a 50% satisfaction deficit for a good hour. Something I've never done before, as I usually tackle the declining power before it happens.
Because of this, the vast amount of water required to keep it running forced a major upgrade to the number of water pumps required. But also shows the issue with the fluid mechanics, as liquid travels quite slowly over long distances.
With the way fluid takes time to travel through long pipe lines, I've been in the habit of creating a small island with water just a few squares away from the reactors, lots of storage tanks with fast pumps connected to the outlet to ensure the pipes are maxed. However, with the 50% deficit it was burning the water way too quickly, leaving the storage tanks dry in just a few minutes and then naturally killing the steam. Were it not for the water being so close by, it would have been a pain in the ass to keep the pipes filled where it needed it most.
Layout
With a sufficient water supply finally satisfied even at a severe deficit, the reactors never ran dry and the steam remained at a stable 3k the entire time. 
So if I were to make a guess, your reactors are too far away from a good source of water and the pipes aren't filling fast enough due to this. Fast pumps help fill the pipes, but they're also affected by electricity and length as well. Keep an eye on the water level in the Cooling Tower Water Source Tank, as a stable flow will keep the water level above 100 at all times, which is necessary to prevent drying out.
Hopefully this is the root cause of your issue, otherwise I'll leave it to the man himself.
(It has now been 2 hours as of writing this and it's still going strong, now I understand why people complain about how much water these things can burn when they suddenly suffer great strain.)
