A narrower follow-up, since the general answer is now clear:

What would you measure differently if you were starting again?

anders_CPH said:

For anyone reading later: the numbers in this thread are worth checking against a primary source before you act on them, including mine.

This is where I part company with the consensus forming above. Filing an objection on the n=1 problem. Personal experience is genuine evidence about one person and very weak evidence about anybody else, and this thread keeps promoting the first into the second.

anders_CPH said:

For anyone reading later: the numbers in this thread are worth checking against a primary source before you act on them, including mine.

Correct as far as it goes. The part it does not cover is what to do when the honest answer is "not enough data", which is more often than anyone likes.

That is the short version; the long version is somebody else's post.

anders_CPH said:

For anyone reading later: the numbers in this thread are worth checking against a primary source before you act on them, including mine.

Genuinely useful, thank you. I had the facts and not the framework. Adding it to my notes with a link back to this thread.

Closing the loop on my own question.

Went back through my own correspondence with an earlier supplier and the warning signs were all there in the replies, weeks before the quality changed. I was reading the product and not the behaviour.