The parallel on-the-fly Crystal algorithm is a new, efficient global search algorithm for exploring single-state potential energy surfaces and conical intersection seam spaces of a wide range of molecules. Despite major developments, its application to complex molecular systems, especially in the condensed phase, remains challenging due to the high dimensionality of the configurational space. In this work, we address this challenge and extend its applicability to the reaction discovery of large and complex molecular photoswitches in various molecular environments, including in the condensed phase with explicit solvent molecules. This is achieved by performing an explicit exploration of a comparatively large Crystal configurational subspace, while gradually relaxing the remaining degrees of freedom. The new Crystal algorithm is applied to the reaction discovery of bilirubin and donor-acceptor Stenhouse adducts, a next-generation class of molecular photoswitches, in vacuum and in the aqueous solution. To this end, we designed an automated and systematic workflow for Crystal to discover and characterize new minima and low-energy reaction pathways in these challenging and complex systems. Our findings demonstrate the algorithm's effectiveness in quickly exploring the configuration space and uncovering kinetically accessible products, offering new insights into the intricate chemical reactivities of these molecules and the roles of molecular environments on the reaction pathways. The results underscore the promising potential of parallelized global exploration methods for reaction discovery in biomolecular systems.