Core library was giving serious blow out of execution speed as data file size increased.

Traced it back to a GetHashCode implementation that was giving too many collisions for unequal objects, so when used as the key in a hash table it was causing the lookup to fall back to checking Equals (much slower).

Improved the GetHashCode implementation, and also precalculated it on construction (they were immutable objects), and run time went to warp speed! Was very happy with that.

Obviously put in a thread sleep to help manage expectations with the boss/clients going forwards. Can’t give those sort of gains away in one go. Sets a dangerous precedent.