Another way to slow down the rate of change in data transfers is to explicitly under-relax the transfers, as described in Under-Relaxation Algorithm.

Data transfer relaxation differs from ramping in two details. First, it is applied every iteration, rather than over the minimum iterations, so the final source-side value is applied to the target side only upon convergence. Second, relaxation is applied against the previous iteration value rather than the value from the end of the previous coupling step.

Data transfer relaxation differs from the Quasi-Newton stabilization algorithm in that the relaxation factor must be set by the user. You can think of the Quasi-Newton algorithm as a sophisticated type of dynamic relaxation which varies per-iteration and per-transfer location.