The effectiveness of high-frequency mechanical impact (HFMI) is considered to rely on the existence of compressive residual stresses. To determine when residual stress relaxation occurs, and what the resulting influence on fatigue improvement is, local stress-strain response in as-welded and HFMI-treated weld toes was modelled under different peak stress conditions. Then, effective notch stress analysis was used to correlate these results with available experimental observations. The simulations showed that high stress ratios and compressive peak stresses were critical with respect to residual stress relaxation, as expected. A compressive peak stress of 0.6fy (nominal yield strength) resulted in full residual stress relaxation. The relative fatigue damage calculations and the notch stress analysis indicated, however, that fatigue improvement could be expected even after significant residual stress relaxation. Based on this and previously observed benefit for high stress ratios, an increase in maximum allowable stresses for HFMI-treated welded steel joints is suggested. The maximum stress ratio is proposed to be increased from R = 0.52 to R = 0.7 and the maximum stress range to limit compressive stresses is proposed to be increased from ΔSmax = 0.9fy to ΔSmax = 1.2fy, which corresponds to Smin =-0.6fy for stress ratio R =-1.