[layer8]

The GP comment is not about designing your own concurrency data structures. It’s about the fact that if your higher-level logic doesn’t take concurrency into account, using the concurrent collections as such will not save you from concurrency bugs. A simple example is when you have two collections whose contents need to be consistent with each other. Or if you have a check-and-modify operation that isn’t covered by the existing collection methods. Then access to them still has to be synchronized externally.

The concurrent collections are great, but they don’t save you from thinking about concurrent accesses and data consistency at a higher level, and managing concurrent operations externally as necessary.