It is certain that there are various definitions of 'compression' What is to be considered is the extent of compression. There is a vast difference in the degree of compression between a press fit and an interference fit. In an interference fit, there is a considered overlap of diameters, such that the shroud has to be heated to expand to slip over the inner diameter. On cooling, the shroud shrinks again and exerts high compression on the inner piece. Aluminium is a soft and ductile material. The head tube is also very thin walled. The head set race is hard and strong and several times thicker. On pressing the head set race into the aluminium flange cup, the flange presumable rides up over a broken sharp edge or champer on the entry face of the race. The aluminium is slightly displaced outwards on full seating of the race. (Possibly measurable with a micrometer) How much compression can the thin ductile aluminium ring really exert on a robust steel inner race? Some, but not much !! very very little (compressive force) Previous poster suggests that the aluminium failed first, eliminated critical compression and created a positioned relief for the failure of the race. I maintain that there is no existence of any significant nor critical compression, considering the proportions and materials here. Therefore the aluminium has virtually no influence on the mechanical life of the race. But the integrity of the race has huge force on the tensioned shroud. There was a key statement in the paragraph above for Capricorn: "The aluminium is slightly displaced outwards on full seating of the race. (Possibly measurable with a micrometer)" This explains the burst appearance of the vertical crack in the head tube. Meaning that the aluminium circumference was experiencing stretch when it failed. This also implies that it is a tear and not a crack. I would also like to refer to the most critical info given by JB in his opening post: the cracks coincide exactly in position. When the steel race cracked first it did open and close slightly under continued use creating an immediately adjacent stress vector on the tensioned aluminium further stretching the aluminium sporadically till it fatigued and "tore". Because the circumference is stretched while the race is still inserted, the gap appears severe. There is a possibility that the aluminium may 'close' if the race is removed, but it will not return perfectly because it has a relative low elastic coefficient. It would be interesting to see if the steel ring opens when removed. ha...If you think it will, we have a problem, because it wont. It already could open with the existence of the tear and the total absence of compression, and wouldn't have to be removed for this to occur. Finally the tear turns and follows the circumferential and vibrating plane of the race: as the forces from the road travel through the fork and crown and transverse through into the race and the aluminium it is in contact with. NOTE: This would occur whether there was any welding, or none what so ever(as I propose) around the based of the head set flange cup form. Failures are evidence or indicators of localized stress and do not unnecessarily arise or propagate to non stress zones, but continue to follow the cause. Sorry if I missed something. HeartCoppi2010-02-18 16:33:11