Not all edge is necessarily ephemeral, if you're thoroughly entrenched and have a deep understanding of a market you will usually just outperform your competitors. Take Jane Street's infamously long-lasting (and profitable!) Indian options trade for example. They had clearly put in the work and effort into understanding a market that other firms didn't do to the same extent, and resources and minds who are good at figuring this stuff out are finite.

You're totally right that edge isn't just knowing if number is more likely to go up, as an example since you mentioned faster access to data: some of the best of the best companies will hire meteorologists so that they know how reliable their microwave towers for transmitting data between e.g. Chicago and NY are (and they can lean in or widen their spreads according to how current and up-to-date their information is).

It sounds like crazy stuff to do, but when your data could be up to 10ms slower than you expect due to weather and you're so sensitive to latency you hire FPGA engineers because normal high performance CPUs aren't enough for you, it's not that crazy.

I sometimes feel like HFT is a waste of good talent and wonder what some of the people I've met who work at JS or CitSec could have done in other industries, but at the same time HFT is often the only industry that correctly prices these peoples' minds. Ultimately having a smoother financial system where risk is more correctly priced is a good thing, even if it's not the best thing they could be doing.