I think this point is very under appreciated, thanks for pointing it out. Along with the 'feature' that could kill things faster is general bloat when people k

I think this point is very under appreciated, thanks for pointing it out. Along with the 'feature' that could kill things faster is general bloat when people keep building without knowing what the customer wants. You have to try something, but product testing should be done in such a modular way that most things can be removed if it turns out to be the wrong direction. Of course, you need a base data structure, but features should conform to that rather than constantly extend. One example I'm seeing with a start-up I currently know is that they are building SDKs for multiple languages. Most of the code is auto-generated, but just the time in examples, documentation and packaging is killing them while they don't have customers for most of the languages they're publishing. This isn't just 'code' debt, there is overhead in documentation and management of code.