The parameters for load tests are different between V2.9.1 vs V2.13.3. Are the following parameters in 2.13.3 (left) equivalent to these in 2.9.1 (right)?
They are similar, but not equivalent. The logic is changed, it is not possible to run fully equivalent performance test. Big change is there that in a turn (Test Step Duration, 10 sec) earlier we executed a conversation just once, now we are executing as many as possible (max 10, because convo runtime is minimum 1 sec).
It means in your case, if your chatbot responding very fast (0.1 sec or below), then you will generate 10 times more load on your chatbot. For example if you would decrease initial users from 40 to 4, then the load should be similar.
If your chatbot is responding slower (1s and above) then the generated load is very similar.