We revisit the "failed" filament eruption of 2002 May 27, first studied in detail by Ji et al. We investigate the temporal and spatial relationship between the filament dynamics and the production of hard X-ray emission using spatially resolved high-cadence data from TRACE and RHESSI. We confirm the presence of a hard X-ray source in the corona above the filament prior to the main activation phase and identify a second coronal hard X-ray source, not considered by earlier studies, that occurs under the apex of the filament during the erupting phase when the filament is clearly strongly kinked. We argue that this second source of coronal hard X-ray emission implies ongoing magnetic reconnection in a current sheet formed via a kink instability resulting from the interaction of the two adjacent legs underneath the writhing filament, in agreement with simulation results. The presence of this second energy release site has important implications for models of solar eruptions.