Some observations from a startup that I was involved in prior to the turn of the millennium:
The company had achieved some small success with a tiny number of customers, but not enough to pay the bills on an on-going basis. Scale required appealing to more customers. Our key barrier to success was that the two key founders (CEO and CTO) had a combined emotional age of 6. They did come to hate each other. Mostly did not talk to each other, for days at a time, punctuated by occasional screaming.
CTO did not have the ability to absorb feedback about making the tool easier to onboard customers, with the result that we would lose potential customers out of the sales funnel because it was just too gruesome to integrate the tool into their work flow (this was a complex electronic design
automation tool). No amount of customer engineer or sales feedback could give that message traction with the CTO. Forcing developers to play the customer engineer role during the sales cycle did not result in reducing the complexity of on-boarding, it resulted in non-scalable one-time hacks that made one customer's benchmark work, and nothing that could be generalized. I lay this outcome at the feet of the CTO that allowed this kind of short-sighted engineering to happen on his watch.
The CEO, by his secretive behavior, created a paranoid culture. For the CEO, information was like coinage that he would spend on you when it suited his purpose. He later made good money in the expert witness business, because he was extremely good at keeping his story straight, and remembering to whom he had told what. He tended to avoid long discussions in large groups because that could expose inconsistencies. And he wasn't only stingy with information, he was stingy with money. One of my most famous quotes about him at the time was: "If he needed so spend a quarter on something to make engineering more productive, he would spend a dime." I've never been one to demand a fancy environment, a used solid-core door from a surplus house held up by two used filing cabinets will hold up a monitor as well as any fancy desk, but I want a decent monitor for my daily coding.
Anyway, my take-aways are to avoid paranoid cultures, spend appropriately (neither too little nor too much) on the things that actually matter, and make sure that key decision makers are able to absorb feedback without going non-linear.