In at least two cases, pg-pool creates a referenced timeout:

1. node-postgres/packages/pg-pool/index.js

   Lines 292 to 295 in 3f6760c

   	tid = setTimeout(() => {
   	this.log('remove idle client')
   	this._remove(client)
   	}, this.options.idleTimeoutMillis)

2. node-postgres/packages/pg-pool/index.js

   Lines 171 to 177 in 3f6760c

   	const tid = setTimeout(() => {
   	// remove the callback from pending waiters because
   	// we're going to call it with a timeout error
   	removeWhere(this._pendingQueue, (i) => i.callback === queueCallback)
   	pendingItem.timedOut = true
   	response.callback(new Error('timeout exceeded when trying to connect'))
   	}, this.options.connectionTimeoutMillis)

These timeouts are currently keeping the process from dying naturally.

Use setTimeout().unref() to prevent this.

Unless I am overlooking a need to keep the pool awake?