Not sure how far application methods for ID surfaces have progressed since I got out of the industry, but PTA hardfacing with WC might be an alternative. Done correctly, it shouldn't crack, and NiCrBSi matrix does pretty well in corrosive environments. May be prohibitively expensive though.
Those pictures are pretty conclusive evidence that failure occurred by fatigue. Flat initiation/propagation zone, large ratchet marks near the point of final failure all scream fatigue.
