I find that one drawback of building your software by knitting together existing frameworks or libraries is that their design choices, interfaces, and abstractions are fixed (and typically not in a way that they are going to fit together seamlessly).

This results in the need for glue-code, sometimes lots, and when you encounter architectural problems, you are much more constrained in how you can resolve them.

Sometimes that constraint is more than offset by the time saved writing the code, sometimes not.

Unless it's something really obvious or standard, these days I tend to start off with a "write it myself" mindset, until I really know I need something off the shelf, which is definitely the reverse of how I was when started my career and had a "just use a library" mindset.