For the more complex features, as a platform-phobic engineer, I find I gravitate towards composable solutions and am happy to re-build unsupported features on top of simple tools because I'm scared of getting trapped in a platform that my needs eventually outgrow. It's great if platform creators are proactive about providing some escape hatches or some modularity so its easy to continue on their platform while also adding in features they lack or aren't a fit for my system.
Related to that, I feel platforms have a higher cost to vet prior to adoption than a single purpose solution. This is at odds with most startup engineering work cycles where you need to get something done _today_ for a specific problem and you plan to grow your understanding of the problem & system incrementally over time. I'm not sure how any software/database platform can overcome this problem.
I generally agree with you - there are too many systems that start off as quick productivity and then you get stuck. So there has to be options and composability in a platform. I really agree that you need to have a way to grow your understanding of the system and evolve it over time - not just chase "quick wins".
I am curious if you found any of my suggestions "too platform-y", since it does sound like we agree in broad strokes about ideal architectures.
