Bring in users in a new system in waves.

New users will discover problems you hadn't found yet.

Once you find that bug, you can fix it so future people won't have it.

Each of those bugs has a high chance of "burning" that user, and making them give up with the product.

When many users come in at the same time and hit the same bug, you get O(1) the useful signal to improve, but O(n) users are burned.

You want to have the minimum number of users burned to generate insight about bugs.

That implies having new users come in waves, so you can fix bugs before the next wave.