Sequential service restoration (SSR) is essential to restore critical loads after the distribution system outage through gradually forming microgrids (MGs). However, high penetration of renewables induces low-inertia issues, especially in small-scale MGs, which could impede a large-size load restoration for ensuring the MG dynamic stability. To expedite the restoration process, we propose the interconnection of MGs as part of SSR. A developed SSR model is proposed to generate restoration sequence and time of MG interconnection operation. An auxiliary variable is introduced to indicate the system topology before and after MG interconnection, and then the maximum permissive step load in each MG is quantified. By this means, isolated MGs are formed gradually, and adjacent MGs can be reconnected to improve the primary response when picking up a large load. Moreover, to address uncertainties of renewables, the stochastic approach is used. And an improved progressive hedging algorithm (PHA) is utilized to solve the stochastic model. The original model is decomposed by bundles of scenarios, and penalty factors are adaptively updated to accelerate the PHA convergence. Simulation results show the effectiveness of the proposed SSR method in improving the restoration efficiency of distribution systems and reducing computational burdens.