You work for a global organization and run a service with an availability target of 99% with limited engineering resources. For the current calendar month, you noticed that the service has 99.5%…

Adding N+1 redundancy would increase operational costs and complexity for a service that is already exceeding its availability target, making it an inefficient use of resources.

SRE principles advocate for identifying and eliminating toil (manual, repetitive, automatable tasks) as a primary way to improve operational efficiency. Automating these tasks frees up limited engineering resources to focus on technical debt reduction, new feature development, and other initiatives that provide enduring value, thereby minimizing operational costs and enabling the team to react to business changes without pursuing unnecessary higher availability.

While defining an error budget is crucial, the goal is to *use* that budget wisely, not simply 'minimize the remaining error budget,' which could imply spending it frivolously or pursuing unnecessary reliability when the service already meets its goals.

Allocating all available engineers to the feature backlog without addressing underlying operational inefficiencies or technical debt is a reactive approach that doesn't ensure long-term stability or cost minimization, potentially leading to increased toil and future incidents.