What I don't understand is how would you determine the next difficulty so that you know whether or not you create an anomaly. And more important how could you influence the difficulty? It's still completely random when someone will solve a block, so therefore it would be completely random what the next difficulty would be. You can calculate an estimate of this but theoretically it could take forever. Could you help me, what don't I understand?
You don't need to determine the next difficulty, Anomalies are based on the _current_ difficulty, which becomes evident when you read the source code (and it seems, the OP in this thread even says that pretty clearly). Basically, (anomaly? yes/no) is determined from height_of_block and difficulty_of_finding_THAT_block. Not difficulty of finding the following block.
You can influence difficulty by abusing KGW. If you look into how KGW works, it's entirely based on nTime - block discovery time as a unix timestamp. Thing is, nTime is very flexible - you can specify at least + or - 100 seconds and it will be accepted by the network. There is nothing random about nTime - it's chosen by the miner... and nTime is "when" someone solved the block