What do your tests do?
Lets start with that. Are they UI based tests or just API based?
If they are UI based then where are the browsers being spun off ? Are they being run in
Docker containers (or)
Locally installed browsers (or)
Do you use a selenium grid?
Have you looked at the environment where your tests are running to see if the resource utilisation ( cpu/memory ) spikes when concurrency increases?
There are a lot of things that can stall/degrade execution.
Feel free to add more contexts around each of these areas.