Design the object model and public APIs for a generic, thread-safe object pool: a container that hands out a bounded number of expensive-to-create objects (think database connections, parsers, or large reusable buffers) and takes them back for reuse instead of creating a fresh one every time. A caller borrows an object, uses it, then returns it; the pool keeps idle objects ready so the next borrow is cheap.

Operating context. A single in-process library used by many application threads on one JVM/runtime. The pool is parameterized by the type of pooled object and never assumes what that object is. Object creation and destruction are delegated to a caller-supplied factory. The pool enforces a maximum size; when all objects are checked out, a borrow either blocks up to a caller-specified timeout or fails. Objects are validated before being handed out and reset to a clean state on return so no caller ever sees another caller's leftover state. Dozens of threads may borrow and return concurrently.

Out of scope. The concrete pooled resource itself (you design the pool, not the database driver), any network or I/O protocol, persistence or metrics export, distributed/multi-process pooling across machines, and eviction of long-idle objects on a background timer (single-node, in-process only).

What to produce. The class hierarchy and interfaces (the generic Pool contract, the pluggable object factory, and any supporting value objects), the public method signatures each type exposes, and the object lifecycle states with their transitions. Be explicit about: the borrow / return protocol and the object's states (idle, in-use, invalid, destroyed); how the factory creates, validates, resets, and destroys objects; how the maximum-size cap plus blocking-with-timeout is enforced under contention; what happens when a borrowed object fails validation or is returned broken; and how a caller that never returns an object is handled.