Elastic systems that are spatially heterogeneous in their mechanical response pose special challenges for molecular simulations. Standard methods for sampling thermal fluctuations of a system's size and shape proceed through a series of homogeneous deformations, whose magnitudes can be severely restricted by its stiffest parts. Here we present a Monte Carlo algorithm designed to circumvent this difficulty, which can be prohibitive in many systems of modern interest. By deforming randomly selected subvolumes alone, it naturally distributes the amplitude of spontaneous elastic fluctuations according to intrinsic heterogeneity. We describe in detail implementations of such "slice moves" that are consistent with detailed balance. Their practical application is illustrated for crystals of two-dimensional hard disks and random networks of cross-linked polymers.