Working with executives of various levels for most of my career, I've learned that smart people trade favors all the time, and are good enough at reciprocity and backscratching that nothing about it has to be said out loud.
This is probably most common in high-end sales; someone signs off on a large purchase, and later, often years later, something good happens to them. A job at the vendor, a position for their kid, luxury conference trips. Nothing that's in the contracts, because both sides are smart enough to know how it works. The space for disguising these favors as normal business is so vast that backscratching is almost risk-free, unless you deal with fools.
It's nearly invisible from the outside, and pretty easy to rationalize from the inside too. Nobody thinks they’re being bribed, rather you’re "doing good work with people you’ve worked with for years". It gets called "trust". The rare cases that get prosecuted seem to mostly involve people who got too greedy, or didn't have enough trust or patience to keep it on the low.
There are some studies that point the same way. Tabakovic and Wollmann (2018) found that patent examiners were more lenient towards firms that later hired them. Physicians who accept even small gifts such as a meal from a pharma company also tend to prescribe more of its drugs.
How worried should we be about backscratching in AI safety and EA? I don't know. There are some vulnerabilities: a culture of trusting quokkas, and a few people directing large sums with wide discretion amongst a small pool of smart people. Outcomes are hard to measure and a favor can easily be indistinguishable from an honest bet that failed. As far as I know, not much acknowledgment that backscratching can come naturally to humans. Being smart and well-intentioned is not much of a protection against this either: in my experience those people are the ones that do reciprocal favors most smoothly.
I'm not sure what can be done about backscratching, though I suspect a reasonable argument could be made about treating grantmakers a bit more like sacred monks, who accept more stringent monitoring of their life and later decisions and forgo some benefits entirely because they could be favors.