When should you use a Docker volume versus a bind mount, and what are the trade-offs?

I'm trying to decide how to persist data for a database running in Docker. I've read about both named volumes and bind mounts, but the documentation doesn't always make the practical differences clear. If I mount a host directory directly into the container, am I risking data corruption if the container writes files while the host process also accesses them?

On the flip side, volumes managed by Docker seem convenient, yet I've heard they can behave differently across platforms—especially when moving from Linux to macOS or Windows. For an interview scenario, how would you explain the portability and performance implications of each method? Is there a default choice you'd recommend for stateful services like PostgreSQL or MongoDB?