Based on my work in software systems yes. Generally when I write a postmortem there is a long list of modifications that could solve the problem. However they

Based on my work in software systems yes. Generally when I write a postmortem there is a long list of modifications that could solve the problem. However they all have different costs, different effectiveness, different specificity and may overlap in different ways. This means that the high-cost, highly-specific modification may not be worth implementing when a lower cost and very broad modification can cover that same class of problems (even if not quite as effectively). As an example you may have two recommendations: 1. Queue the requests and add many retries so that if a problem occurs in the future the requests are not lost and the problem can be fixed or allowed to resolve itself. 2. Add a second implementation of the process using different technologies and dependencies and fail over to it if the first implementation fails. If this specific type of request is not very latency sensitive 2 is likely not worth it even though it is an improvement over just doing 1. 2 might be the better solution as well however 1 would greatly mitigate not only the specific request that had a problem this time but all similar requests in the system so 1 is good to do even if 2 is done for the other types of request. In the end I would say that both are good recommendations however the logical business choice is just to do 1.