upvote
More on technical side I'm interesting what is plausible explanation for this type "glitches"?: it inconsistent backend router state between processing nodes, processing application restart and screw up in shared memory segment (i can imagine to decrease load times - use "persistent" shared memory block for outstanding data), or just plain hash table collision and lack of empty slots (i mean: https://en.wikipedia.org/wiki/Hash_collision).
reply