One of the main parts of being a scientist, at least as measured by time spent, is writing grant applications. Scientists spend a huge percentage of their working hours asking for money from an elaborate, high-overhead funding system that goes to great lengths to try to avoid spending money on the wrong things. But in practice the correlation between attractiveness of grant proposals and the quality of the resulting research doesn't seem to be strong, meaning that this process might just be, at least to a first approximation, an ever-more-expensive random number generator.
You can certainly imagine other ways to do it. Funding agencies could try allocating money at coarser granularity, e.g. larger grants shared among more researchers with much less specificity about how the money is to be spent; I hear this used to be more common many decades ago. Or they could do a quick-and-dirty screening for what 1-2 page research proposals aren't obviously terrible and send out money to the vaguely plausible ones at random [1], which would at least have lower overhead. Or you could have grant money allocated by individual people instead of committees, with less deliberation and guaranteed fast turnaround time to a yes-or-no decision, e.g. [2] and [3].
In general there's usually more than one way to do something, and I worry about a monoculture getting stuck doing things in a deeply suboptimal way but unable to change. I could tell similar stories about the modern peer review process (a relatively recent thing, and arguably not very good at solving the problem it's meant to solve), or the persistence of bad statistical methods in a lot of fields because they're the standard that everyone expects, or various other things.
[1] https://www.vox.com/future-perfect/2019/1/18/18183939/scienc...
[2] https://marginalrevolution.com/marginalrevolution/2018/09/em...