Since its invention twenty years ago, photonic crystals show great promise for full control of light propagation and emission. Tremendous progress has been made in design and fabrication of novel photonic crystal devices and circuits. Unfortunately, widespread application of photonic crystals is often limited by light scattering by structural disorder or surface roughness which is unavoidable in the nanofabrication process. Such uncontrolled random scattering is shown to be detrimental to passive photonic crystal devices such as waveguides, as it contributes to optical losses and limits light propagation length. In the visible and ultra-violet (UV) frequencies, the effect of structural disorder is more pronounced due to small characteristic size of structural features. The question we address is how the random scattering would affect the performance of active photonic crystal devices, e.g., photonic crystal laser. Contrary to common expectation, we illustrate through simulation and experiment that the inevitable disorder is not detrimental to UV photonic crystal slab laser made of ZnO. Under certain conditions structural disorder may lead to spontaneous optimization of optical confinement in a photonic crystal slab by automatically balancing the in-plane and out-of-plane leakage rates. We attribute this counter-intuitive effect to the fact that optical gain selectively amplifies the high-quality modes of the passive system. Despite the disorder, the photonic bandedge effect enabled us to efficiently extract optical gain and to fine-tune the lasing wavelength from 383nm to 407nm with sample-to-sample fluctuation of about 5nm. Because our approach to designing microlasers does not require flawless nanofabrication and carefully-designed structural defects, it is expected to result in inexpensive laser applications operating at visible and UV wavelengths. [A. Yamilov, X. Wu, X. Liu, R. P. H. Chang, and H. Cao, Physical Review Letters 96, 083905 (2006)]