Check the tolerance value and the maximum change for the last iteration.

Did the model truly "not converge" or was the tolerance unreasonably small? A good mass balance would suggest the tolerance may be too small. Alternatively adding a few more iterations may cause the model to meet the tolerance easily.

Study the progress of the solution with each iteration and either increase the maximum number of iterations, adjust the relaxation parameter, or increase the tolerance if appropriate.

If these actions are not productive, consider the cause of the problem.

Is there one cell or a number of cells in one area that always reflect the maximum head change? If so, carefully check your input in that area, there is often a "typo" yield a large contrast in properties of adjacent cells.

Does the problem make "hydraulic sense?"

Is there a very low gradient in some area of your model that will require a tight tolerance and many iterations? This should also be reflected in a poor mass balance.