Really, the choice between SQL and NoSQL systems depends on what kind of system you're running. SQL is relatively expensive compared to most NoSQL systems because it provides all the ACID guarantees -- Atomicity, Consistency, Integrity, and Durability. These are important guarantees to maintain consistency of your data if you actually require consistent data. If you don't require consistent data (e.g. you are a caching solution or you are Twitter) then NoSQL systems' efficiency becomes much more attractive.
For your specific use case; it doesn't sound like there are many solutions that are going to help you. Modifying the middle of a text file is inherently going to require (at least) rewriting the entire portion of the text file after the edit point out to disk (assuming you actually want the files to be plain text on disk).
You might be able to build a system on top of SQL or NoSQL which represents text files as lines or chunks of lines, and be able to operate on them in a row-oriented manner. But even that type of system is likely to be inefficient for files as large as 100-600MB. Consider storing the files themselves as some kind of structured data in SQL; and then regenerating the files on demand when a user requests the full text file.