Dr.EndoEP said:

It is worth asking what the claim would look like if it were false.

Adding the part of the answer the thread has not reached. Whatever the answer turns out to be, the method for getting there is the same: state the assumption, do the arithmetic in public, and invite the correction. That is slower than asserting, and it is the only version that survives being wrong.

I would rather be corrected than agreed with, if it comes to it.

Adding the numbers, since they settle part of this. Say what you would expect to see if you were wrong, before you look. It is a small discipline and it changes what you notice.

I would rather be corrected than agreed with, if it comes to it.

Following on from Dr.EndoEP — and this may be the naive question:

How long did you give it before you decided it was working?

raj_cambridge said:

Say what you would expect to see if you were wrong, before you look.

I am going to disagree with the direction of travel here. The thread has converged on the tidiest answer rather than the best-supported one, and those are not the same thing.

Closing the loop on my own question.

Update: the thing I was blaming turned out not to be the cause. Leaving the original post as written rather than editing it, so the mistake stays visible for whoever searches this next.