Cellulose is the most abundant biopolymer on the planet, with significant potential as a feedstock for bioderived products. In nature, fungi and bacteria secrete cellulases to hydrolyze cellulose into smaller carbohydrates for incorporation into their own metabolic processes. For effective biomass utilization at industrial scale, these cellulases are coupled together to produce a clean sugar stream. One such cellulase is Cel7A from Trichoderma reesei, which is used extensively in biotechnological applications. However, dissociation of Cel7A from a bound cellulose strand is rate-limiting, reducing catalytic efficiency and yields within industrial applications. To explore this unknown dissociation mechanism and devise potential enzyme engineering strategies to mitigate slow dissociation, we conduct Hamiltonian replica exchange molecular dynamics simulations. Two postulated dissociation mechanisms were tested ('dethreading' of the bound cellulose strand from the binding tunnel and 'clamshell'-like opening of the binding tunnel and recrystallization of the bound cellulose). For each mechanism, free energy profiles were determined together with estimated diffusion coefficients. The kinetics for both mechanisms were then estimated and compared against experimentally measured dissociation rates, which suggest that the dethreading mechanism ought to be preferred. These findings are further supported by dissociation rates measured on engineered mutants incapable of undergoing the clamshell mechanism.