Chunking mechanisms, the processes of grouping several items together into a single processing unit, are central to several cognitive processes in human and non-human primates and notably to the acquisition of visuomotor sequences. Individuals segment sequences into chunks to perform visuomotor tasks more fluidly, rapidly, and accurately. Using an operant conditioning device, we previously studied the precise mechanisms by which chunks are formed and reorganized during sequence learning. Eighteen Guinea baboons (Papio papio) repeatedly produced the same fixed sequence of nine movements during 1,000 trials by pointing to a moving target on a touch screen. We found that chunking patterns are reorganized during the course of learning, with chunks becoming progressively fewer and longer. We also identified two forms of reorganization of the chunking pattern: the recombination of preexisting chunks and the concatenation of two distinct chunks into a single one. To understand the conditions under which these reorganizations occur, we study here how the stability of a chunk and the stability of chunk boundaries are related to these reorganizations. Our analyses showed that less stable chunks and less stable boundaries are more likely to produce reorganizations. These results provide new evidence about the fine-grained dynamics of chunking mechanisms during sequence learning.