A related opinion, nothing on switching between 6X and 12X.

We can argue till the cows come home to roost what approach a manufacturer should take when an upgrade happens.
Over the years, I have observed the Philips approach to "upgrades" and I think The Philips philosophy is "let us not lose the advantages of the new design for the sake of the ultimate compatibility.
I think the most dramatic example is what happened when Philips and Intel broke up their collaboration on a "16 bit '51". Intel went through hoops to make the 251 object compatible and came out with a somewhat improved product. Philips designed the absolutotally wunnerful XA and made it "source compatible". It takes 1-3 days to convert a '51 assembler project to run on the XA in '51 mode (I would not recommend that approach) but if you go "C source compatible" it burns the asphalt leaving the 251 in the dust. If then you take the next step and utilize the features such as hardware task switching you end up with something far superior. Ok so it is not fully "compatible" to a '51 but should the desire to be that make you end up with an inferior new product.

Of course we all hated replacing the supervisors when the 51H became obsolete, but after the pain had ceased I hope we all agreed that it was worth it.

OK, so I like the way Philips make new stuff (have a look at the 89LPC932) and some of you like the way Atmel does it; well let us agree to disagree.

I really do not think one 6X/12X approach is superior to another, both do the job,

Erik