You both talk about the perils of dynamically building queries, yet proclaim the benefits of static syntax checks. You can't "statically syntax check" a dynamically built query. Even with ORMs. So pick one. Either you're building it dynamically, or if it's static there won't be "string pieces" to worry about.
Most of queries are defined statically: a query with several anonymous (?) or named (:foobar) parameters to insert literals. So none of those scary string concatenation problems even occur.
Regarding stray commas... please: listOfIdents.join(','). It's not rocket science. Plus, I'm not necessarily advocating you use string functions to build queries, there are many small helpers to be had when you need to build a complex query, but the point is to know when to stop, and not try to completely hide the resulting SQL from the user of such libraries.
It's one thing to use a query builder for those rare complicated dynamic queries, but an entirely different thing to lock down and hide the connection and have your hands at the mercy of an abstraction layer for your every query.
If you can't control when and how your database gets called, if you can't control select locking, and transaction isolation levels, if the vast majority of its functions and syntax is out of reach, you can only do the most basic of CRUD operations on SQL.
It's just ugly that we have a complex piece of machinery like a modern SQL database, with all those advanced querying and MVCC features packed in for people to use them in order to build fast and safe applications, and watch them use it like a little more than a file system.
It's like buying yourself a million dollar supercomputer so you can play Solitaire on it.