Due to the aging of structures, the issues of plant life management and license extension are receiving increasing emphasis in many countries. Understanding failure of structures due to random roughness on surfaces at early stages of degradation is therefore crucial. It has been shown that even slightly sinusoidal roughness can increase stress concentration by a factor of 2 or 3, which can be critical for a brittle component due to the significant reduction of its load-carrying capacity, even with slight roughness. A more in-depth fracture analysis of surfaces possessing random roughness is needed in order to more profoundly understand, and hence develop models that will predict more accurately, failure of structural materials exposed to degrading, in-service conditions. Using a technique previously developed and successfully applied, replicates of random rough surfaces, imprinted with various levels of degradation, and at three distinct auto correlation lengths, were realized and mechanical testing was performed on them. The stress, strain and energy at fracture are reported. Finite element analysis was carried out to elucidate experimental results. Besides the expected reduction of energy at fracture with degradation, a relaxation region was observed where the energy slightly increases. This phenomenon implies that even after degradation has progressed there is a local maximum of energy at fracture due to the competing effect of tendons and growth of pits. The results find applications on the early stage of maintenance of surfaces of structures in service.