Luckily we were still taking between 30-50% of normal traffic for the full 3 1/2 hrs, when by the book after 1h5m the whole airspace should be cleared and MST officially closed except to state and hospital flights.
From what we here the failure was somewhere in the LAN connecting the various servers; the pri. system was all over the place but the back up system worked very well. In fact the only reason we were able to stay open is the fact that the back up system in MST is now very advanced.
It also happened at one of 'those' moments, planes all over the shop due to CB avoidance, a real heavy workload for a while.
The fact that the failure lasted 3 1/2 hrs is quite extraordinary; we were in fact in the final moments of switching back to the pri. system after about 20mins when it failed again, and so for obvious reasons stayed on the fall back for sometime longer
