It suggests that the complexities encountered with merging, such as non-commutativity and non-associativity, often arise by chance due to the accidental ordering of transactions. The point made is that while it's impossible to completely eliminate these random occurrences, it may not be worthwhile to invest time in exploring additional subsets for potential improvements. This is because the probability of discovering a significantly better solution is low. Instead, the recommendation is to allocate efforts towards experimenting with more options within the linearization algorithm itself, which could lead to more productive use of time. This approach underscores the value of focusing on direct enhancements to the core algorithm rather than attempting to perfect peripheral aspects with diminishing returns.