I thought these quotes from well known racing guru/author Carroll Smith are pertinent here, from his 1990 book "Nuts, Bolts, Fasteners and Plumbing Handbook", pages 26 &27:

Regarding Fatigue:

"The initial crack has reduced the cross-sectional area of the member. Consequently there is less material available to resist the next load cycle. We know that unit stress is load divided by cross-sectional area. Therefore, the next time the load is applied, the imposed unit stress will be greater than it was the last time (assuming an equal load). More important, despite the migrated dislocations, the jagged bottom of the initial crack acts like a first class stress concentrator. Just as wolves always go after the weakest member of the herd, as more load cycles are applied, the concentration of stress at the leading edge of the crack will cause the crack to enlarge until enough fresh material is engaged to resist the stress, this time. These fatique cracks are transgranular in nature; the fracture actually splits the individual grains of the metal rather than following grain boundaries. As a result, the opposing surfaces of the crack tend to be quite smooth in appearance."

"This repeated sequence of events creates typical smooth opposing surfaces with the telltale concentric beach marks that are also visible in the illustration. The beach marks are formed by the progressive enlargement of the crack and radiate outward from the focus of the original fault. Beach marks are characteristic of the fatique failure...…… The next sufficiently large load application sudden and catastrophic failure of the remaining portion of the metal. This last failure is intergranular in nature."

"The final rupture proceeds along the grain boundaries, leaving the individual crystals exposed as a rough and granular surface. This allows the self appointed expert to peer knowingly at the part and proclaim, "Ha! Just as I suspected, a crystallization failure" . When you hear this sort of statement, the proclaiming person from your list of those to be consulted in the future."

"it is important to realize that, in the fatique history of any part, neither the rate at which stress is built up within the part nor the period of the time over which the stress is maintained is significant."