We were actually not solving the exponential complexity problem on TypedPipe equality. With this, we have a test that it was exponentially slow, then fix it.
Without this, applying rules is exponentially slow, which is obviously no good.
In the long run it would be nice for the timing/algorithmic complexity testing code to be pulled out in a more generic place so we could use it other places.
We were actually not solving the exponential complexity problem on TypedPipe equality. With this, we have a test that it was exponentially slow, then fix it.
Without this, applying rules is exponentially slow, which is obviously no good.