Hamiltonian constraint in Euclidean LQG revisited: First hints of off-shell Closure