You are right when you said that Intel handled the 900 MHz Xeon recall correctly. They acted swiftly and decisively by stopping production and recalling the parts that had already been shipped. I give Intel many kudos for the actions they took when they realized there was a problem as well as the quality control team that found the problem in the first place.
However, I think you went a bit to far when implying that the 1.13 GHz P3 recall was also handled properly for two reasons. First, the 1.13 GHz P3 should never have been released in the first place. Intel essentially had to overclock the chip to get it to that clockspeed. Second, they didn't find the issue on their own (or even worse blatantly ignored it) and didn't issue the recall until the proverbial hammer was brought down on them by several hardware sites led by Tom Pabst of tomshardware.com
As for Memory Translator Hubs (MTH), to paraphrase from an old Woody Woodpecker cartoon, "If Intel hadn't tried to force Rambus down our throats, this would never have happened". Intel learned a hard lesson at that time. The x86 enthusiast community are not like Mac enthusiasts. They do not blindly accept what any manufacturer releases as being wonderful just because the head of that manufacturer says it is.