well, what happens is that as i make refinements to the dims and parts of the cube (eliminating unused attributes, looking for rigids vs flexibles, create a new calculation, drop a hierarchy etc etc etc) it works swimmingly, but when those things change in the cube concurrent with the data warehouse refresh, things go awry. i have succeeded in getting it to work by reprocessing each dimension separately, then each partition, then the cube all together. so my intuition about whats going on is that i feel that some parts of the fact are getting processed ahead of some dimensions that should have been processed before it, or whose processing should have completed by the time the facts get started, because im at a loss to explain why doing each operation manually (and serially) 'fixes' what's happening. some of the failing dimensions have as many rows as the fact (e.g. claim_utilization_fact has as many rows as claim_detail_dim...when i process the dim separately and then process the cube, it finds its keys <g>)