The reason this (splitting the argument into many small pieces) is a problem is that sometimes, checking an argument by going over it with a microscope isn't sufficient.

If the checker is mostly reliable but fallible, a long-enough inferential chain can defeat any specific amount of reliability, making the checker useless at best and maybe even deceptive.

However, even though it is a flaw, it's not a fatal flaw. We can strongly prefer smaller arguments, and maybe there are other techniques too (representing is-a-refinement-of explicitly?).