Tungsten is the privileged option as plasma-facing material for the divertor region of ITER and DEMO tokamaks. Under repeated high thermal fluxes (10 MW/m² in steady state and 20 MW/m² in quasi-steady state), Plasma-Facing Units (PFU) can be damaged through thermal fatigue phenomena, which are enhanced by in-service tungsten recrystallization. Slowing down tungsten recrystallization is thus a way to optimize PFU lifetime. Tungsten-recrystallization kinetics is known to depend upon the initial microstructure. In the present paper, a mean-field model is developed to highlight microstructural effects upon tungsten restoration mechanisms at high temperature. The effects of grain size, initial recrystallized fraction, initial dislocation density, i.e. stored energy, and dislocation density distribution have been investigated. It is clearly quantified how a higher stored energy can lead to a drastic increase of the recrystallization rate. It is also shown how a prior recovery stage helps to slow down recrystallization. Finally, a conclusion is brought upon the ability of such a mean field model to optimize the tungsten initial microstructure for fusion engineering.