Maybe it's just the way I use debuggers, but I don't find this to be a particularly compelling reason; I use instruction-level debugging quite often even when I have the source (debugging a program I wrote, when simpler methods fail), to [1] verify that the high-level language statements are doing the expected, [2] ensure that the compiler is generating the expected instructions, and [3] to observe where exactly the bug is.

> IMO gdb isn't very good for binary-level debugging

Having come from a Windows background, I'd say using gdb for that is an absolute nightmare - it appears to almost be designed to actively make it harder to do instruction-level debugging, despite the fact that the implementation of a debugger for source code is far more complex than one that doesn't. E.g. the "disassemble" command absolutely requires a symbol table, and will still complain if there's no "function" at some address, when all it really needs to do is start reading the bytes at the address it's given as a parameter - which could be the name of a symbol - and interpret them as instructions. Instead the somewhat unintuitive "workaround" is to use "x/i".