We recently had this issue, but the Connection Pool Error message was eventually just baiting.
The real problem was a third-party system that did not accept more messages, thus blocking each thread. The connection never returned to the Pool if the thread had previously acquired a Connection from HikariCP. Hence, we had as many successful connections as Pooled Connections.
In such a case: Inspect the number of blocked Threads on the JVM with a tool of choice (any APM, VisualVM, or alike), and inspect where the blocked threads are halting with a Thread Dump.