Knowledge workers aren't cogs, they're part of a team. You shouldn't expect that you can easily find replacement people with exactly the same skillset and propensity as someone else to make up the gap that an absence makes in a team. Nor should you expect that the proper way to grow a team is by cloning the skills of some other existing member of the team. A team works cooperatively, and their skills, talents, experience play off one another to gel together into some sort of mechanism that is capable of doing stuff. But if you change the parts, you get a different team that does different stuff, or maybe doesn't even work at all. One of the big factors here is that if you think you can replace one person with just another person you're often wrong. Best case scenario you end up with something else (maybe better), worst case scenario is you can't replace the unique factors that made the previous person successful in that role (and realistically you need a team with a different breakdown of skills, different mechanisms of interconnectedness, etc. in order to have something functional).
The band analogy really helps here a lot. Realistically when you change members of a team you don't end up with something like Steve Perry stepping into the lead singer role of Journey. What you end up with is something more like Fleetwood Mac being transformed by merging with Buckingham Nicks or Genesis changing after Peter Gabriel left. You end up with a very different thing doing very different stuff. And the higher caliber the people the bigger the difference is.
What knowledge workers actually do day to day and what it says on their job description they do are often very different things, and you ignore that at your peril. Hiring based on the idea that knowledge workers are automatons with certain functionality modules installed is one of the easiest and most common ways to completely sabotage productivity and execution.