That number represents the total number of processes across all three nodes. The 1000/1000 is a limitation per node and it is a soft limit, not a hard limit.
FYI, 1000/1000 for Maximum Timer Driven and Maximum Event Driven threads are unrealistic values. The Maximum Number of Timer Driven threads should be between 2 and 4 times the number of CPUs on the nodes. The Maximum Number of Event driven threads should remain at 5.
In addition to the above correct answer. HDF 3.0 has a bug where the max Event Driven thread pool shows as active threads in the UI. With 1,000 set for Max Event Driven and 3 nodes in your cluster, that accounts for 3,000 of your 4,958 there. Also, if you increase the Max event Driven thread count and then decrease it, the active thread count will still show the higher setting. Only a restart of NiFI will trigger a reduction.