Cloud manufacturing service composition reconfiguration (CMSCR) is an essential process for handling unpredictable service exceptions to ensure the smooth operation of the cloud manufacturing (CMfg) system in a dynamic environment. Considering the occupied status of service providers of a CMfg system (CMSPs) at the time of change occurrence, the reconfiguration can be organized only with the available time of CMSPs, i.e., a set of available service time windows (ASTWs). In traditional CMSCR studies, tasks are assumed to be processed in fixed size of batches, which will lead to the unavailability of some ASTWs. This inevitably results in the insufficient utilization of CMfg resources and leaves less room for reconfiguration. To handle this problem, we introduce flexible splitting and intermingling strategies in CMSCR, aiming to improve the reconfiguration capacity by increasing resource utilization. This paper first analyzes four typical types of unexpected changes in ASTWs and their response conditions for reconfiguration. Next, an enhanced CMSCR approach with flexible splitting and intermingling strategies (SCRTW-SI) is proposed to handle the unexpected changes in ASTWs. In addition, a novel slack-based insertion mechanism is developed to further improve the reconfiguration performance. The CMSCR problem under consideration is formulated with a multi-objective mixed integer programming model. And a multi-objective service composition reconfiguration algorithm based on memetic algorithm (MOSCRMA) is proposed, in which some problem-specific schemes are elaborated. The performance is validated through extensive numerical experiments. Finally, a real-world case is analyzed to demonstrate the applicability and superiority of the approach.