It is all rather odd. The first issue is the fact both MCCB and fuses go, typically if well designed that means a very big fault, i.e. so the MCCB let-through is above the fuses pre-arcing I2t and so even though it trips, the fuse is already a goner.
Now we don't know the MCCB type or settings in any detail here, so that current where selectivity is lost might be as low as 0.8kA or so if bandly designed...but the time duration of the 1.9kA as 2 cycles or 0.04s is not enough to explain the fuse going unless it was already really hot.
However, the currents prior to that were only around 200A so fuses would be effectively cold then?
So did the monitoring get something wrong? Has it some upper limit where the CTs saturate and really the surge was really closer to 5-10kA for 2 cycles?
I guess there might be somewhere that has something causing a major fault current such as water getting on to terminals or similar and it going all arc-flash as typically a L-anywhere arc are migrates so you end up with bolted 3-phase style fault. But surely someone would notice the sound and/or damage? Doing that a couple of times I would expect to burn off metal and make it permanently shorted or open enough to stop.
The OP has not said what time it is tripping at, maybe that would give us more ideas?