You touched upon this, but I think that a big part of the resolution of this question is tackling that we cannot always easily discern what to do. If there is a better option but we can't find it, and in trying to find it we adopt strategies that frequently lead to bad outcomes, we're clearly taking a worse strategy than we could.

A lot of philosophers seem to have preferred not to spend the bulk of their effort considering cases where they do not know the consequences of their actions and have to make the exploration/exploitation-style decision about how much effort to invest in understanding the situation before acting.

But this is where we all live: in the world where we struggle to fully understand how even those close to us will react (and what they need), much less those with whom we're less familiar. We can even have trouble understanding ourselves.

In this knowledge-poor environment, a local optimization strategy, coupled with global customs that prevent the local optimizers from mutual antagonism and tearing down, is highly likely to be among the best strategies.

If you add in that we've evolved for local preference and have trouble actually behaving otherwise (even if we can train ourselves to will otherwise), it turns the high likelihood into an almost-guarantee.

The questions that remains are then: what should the global customs be to allow mutually-productive-or-at-least-not-mutually-detrimental local optimization; and how do you make a good life for yourself and those around you without violating the global customs.