There is a huge pool of exceptional junior engineers

Hiring juniors is always great if you, somehow, have a much better filter for finding the stars than the rest of the market. But if you don't, hiring bad juniors is a disaster: No different than outsourcing bits to a bad satellite office.

So are you actually good at finding the good juniors in this very difficult environment? Can you change your hiring machinery to improve, as most traditional ways have stopped working? Because hiring a lot of juniors that don't work out sure can kill companies.

Hire one junior per team. Don’t overload your senior staff with OKRs and managerial tasks. Let mentorship and apprenticeship happen.

I guess what’s the value of the junior there? Why is that superior to just having the seniors have their heads down coding and not being pestered by a junior?

What do you do when your seniors move on or retire?

Also, even seniors are usually more than happy to outsource work they've already done a million times, but that's still new to the junior ("build the Terraform to stand up this cluster" etc).

People only average a few years in a job these days. The juniors are most likely to end up as seniors somewhere else. So hiring juniors who provide negative value at the start is mostly benefiting the industry as a whole at your own personal loss. Which makes it a pretty easy thing to cut.

I know in other industries they have a kind of lock in where they provide free training under the condition that you work at the same company for a number of years. Which sounds bad but I don't see many alternatives.

Give them decent pay increases and I'll guarantee you they won't leave.

People can leave for a lot of other reasons than money. You can't control or predict that.