After 40 years experience as a metallurgical engineer maybe I can add some info on these failures. One of the key factors in hydrogen embrittlement failure is sustained tensile stress. The generally accepted theory is that monatomic hydrogen atoms dissolved in the steel matrix (classically from plating operations but not always) diffuse towards the larger interstitial spaces in high tensile stress areas and help precipitate brittle behavior of the metal at the stress raiser or crack tip. Thus hydrogen embrittlement is an issue in parts like spokes, QR shafts, and various high stress bolts where there is a high sustained tensile stress. Not so much in parts subjected to cyclic stresses as there is not enough time for the hydrogen diffusion during each tension cycle. There is also a tensile stress threshold below which hydrogen embrittlement is unlikely in steel. Without looking it up, it’s around 160,000 psi which is why it’s almost unheard of in parts with a yield strength (or hardness of 40 Rockwell C) or less; the parts yield and thus do not see high sustained tensile stress. So lots of SAE grade 8 bolt hydrogen embrittlement issues, not so much a problem with SAE grade 5 bolts.

The fracture faces show signs of reverse bending fatigue; note that the fracture features are roughly symmetrical about a mid diameter line. If Dawes-man were to fit the broken stubs back into the same cranks and the mid diameter line was roughly parallel to the ground when the cranks are also parallel to the ground (maximum bending stress position) that would tend to confirm my hypothesis. Bonus points if the “wider” part of the fracture is “up”.

Auchencrow is not wrong per se, but is misinterpreting the evidence. Most (less than half in my experience) fatigue fractures do not display obvious fatigue striations (also known as “beach marks”) as in his (please excuse if you, Auchencrow, are other than conventionally male) photo of a classic bidirectional bending fatigue fracture @ 1/23/11 12:16 am so clearly shows. Many fatigue striations don’t show short of scanning electron microscope examination and even that is not a sure bet. His photos at 1/23/11, 12:35 am are clearly unidirectional bending brittle failures with the fracture initiations at 6:00 O’clock as indicated by the overall direction of the river (or “chevron”) marks. Those marks tend to “V” out from the initiation site. But hydrogen embrittlement fractures being typically brittle does not equal brittle failures being typically hydrogen embrittlement.

If it were my pedals and nuts, I’d
1) Polish the pedal shafts in the area where the fractures are occurring to a 600 grit or better finish to minimize the stress risers and make subsequent inspections easier. Various abrasive papers and cloths make that reasonable even if you have to use them “shoe shine” method. Better if you can run the final polish marks parallel to the spindle rather than circumferential.
2) I’d tear down the pedals every few hundred to a thousand hours of riding (if you ride a couple plus hours a day that’s about once a year) and seriously eye ball the suspect area for cracks. Dye penetrant or magnetic particle inspection would be a bonus but is probably overkill if you’re diligent about looking for cracking. Less so after the first couple of thousand hours. There is a “fatigue limit” for steel parts. Barring impacts, steel parts (does not apply to aluminum ect.) subjected to ten to the seven cycles (10 million cycles) without failure are unlikely to ever fail. 1000 hours of cycling at 100 rpm equals 6 million cycles.