Thick near-wall prismatic layers and large offsets can be constructed using springback procedure when thin highly compressed layer of polyconvex hyperelastic material is attached to fixed surface and is allowed to expand. This procedure allows to construct one-cell-wide layers and offsets with thickness comparable to the characteristic size of the body. Resulting mesh layer does not contain inverted cells. Self-intersection zones can be easily eliminated by cutting off excessive thickness. This procedure generally does not lead to offset thickness reduction due to local dents or elevations which is quite different from the advancing front collision detection technique. When small surface elements are present discrete variational problem can become quite stiff due to very large target height-to-base ratio of expanding elasting cells. Problem with convergence of iterative procedure is manifested as final thickness falling short of target values. Thus iterative minimization of hyperelastic stored energy needs finely tuned preconditioning. Note that the Hessian matrix for discrete polyconvex stored energy is not positive definite and problem of choosing good positive definite preconditioner is not trivial. We present implicit solver for variational springback problem and illustrate its behaviour on hard real-life test cases.