The issue is that many people already have a substantial proof of their coding ability in form of completed work. The rational approach would be to evaluate that work and, as safety, test that they have actually performed it. Whiteboarding them catches nothing other than inability to perform unusual tasks under stress.

As as we know, engineers are never in stressful situations so no point testing it.

Production never goes down, critical errors don't arise and time-pressure tasks are rare.

People do not behave consistently under stress so such a testing value is close to 0. It's a fight or flight response. If "fight" rolled up on the interview there is no guarantee it will roll up next time your production goes down. You can just as well toss a coin.

...and yet soldiers in WW1 and WW2 were subjected to stressful situations lasting years, some of which manifested in mental trauma but many of which actually behaved in the manner required.

My point is; you are neither psychologist nor professional human resources executive search.

Your reasoning for not conducting testing interviews is flawed and can be challenged at every juncture precisely because this is not your area of expertise.

It's not mine either but I am not advocating sweeping policy changes across an entire professional landscape.

If you feel there is a better route; then write up, test it, get it peer reviewed and implement it.

You can certainly challenge anybody's reasoning just as well as you can challenge anybody in court over any matter. But it requires more than typing first thing that comes into your mind for such a challenge to stick, jfyi.

But not in this case, the first thing that came to mind was a sufficient rebuttal.

Since you are constantly referencing the legal system you should remember it is the prosecutors role to prove the hypothesis. It is not on the defence to disprove it.

Merely rebutt sufficiently.

You have not made a single compelling argument for not testing coders with the current model.