Each optimization that a solver performs can be associated with a single solution path.
A solution path is characterized by the sequence of steps and decisions the solver makes as it performs the optimization. A model can have a single optimal solution, yet there are many solution paths that the solver could take to arrive at this solution. Alternatively there may be multiple optimal solutions and different solution paths may lead to different optimal solutions.
Under certain conditions Gurobi is deterministic, meaning that the solution path will be the same each time the optimization is performed. These conditions include the model itself and the value of Seed. Changing the value of Seed is the easiest way to cause the solver to take a different solution path. The variation in results arising from different solution paths is known as performance variability, and is an important factor in benchmarking and making accurate comparisons.