Respinning PCB's, for example, involves risk. If I have boards made here in the U.S. I'm fortunate to have a 1% yield by the time they're assembled.
You need to change something, my US made boards yield somewhere above 99% for production, prototypes (a low count) has yielded 100%. I will gladly let you have the names of my contract manufacturers, send me an e-mail

You see, things look pretty different from our quite different situations.
Of course they do. If you make only a few, the engineering cost is the concern, when making many unit cost is. The problem I have is not the statements you make, it is the "universalism" they are stated with.

I would never have a problem with a post stating: with "when you are making only two units ...." but that is NOT what you say, you say "using the internal UART is silly" without any context, as if it is globally true.

I think you get into the position of assuming that someone, anyone, doing something differently than you would do them is WRONG. I don't believe that.
Neither do I, there is more than one way to skin a cat.

That's why I jump on you every time you tell someone they shouldn't use the i8255, though I wouldn't do it unless the thing were already on a board, or, that they should use your favorite MCU, which is seldom what the person making the inquiry wants to do.
I have never "jumped upon" someone that had to rescue (not reuse) some old board with a 8255, I DO consider it ridiculous, however, to use it when you start from scratch. As to "my favourite uC" I have suggested LPC, 66x, rx2, SILabs and if that is A favorite uC I must have misunderstood something. If you look at my uC "rcommendations", you will see they are application specific. Yes, I do not recommend your favorite, but you can not find me stating anything against it, I just have bypassed that step and gone on. When the choice was 12 clocks or Dallas 4 clock, I actually did use them in some apps, but that is so long ago, that I can now neither encourage or discourage their use.

Now, I don't think disagreement is a bad thing in itself, as it reflects differring views of the situation. I find I get much valuable information from folks with opinions that differ from my own, but its value does depend on why it is that you disagree.
As stated in the first paragraph "not the statements, the universalism". If you read my posts here I have said e.g. "when using mode 0..." "many small derivatives do not have a T2" which show that I may agree with what you say if the context is stated but stating as a general rule that a baudmatch crystal is not needed when serially communicating is outright false. If you had followed the various posts here, you would have observed that there is not one single post that state "I need RS232 for debugging purposes" and posts "I need RS232" are legio.

Erik