让我解释一下我的设想。
我有一个包含550万条记录的简单表，结构很简单。

relations
------------
id
parent_id
name

我有一个递归函数来获取一个节点的所有子节点id;这个函数在任何情况下都能正常工作，这不是问题所在

CREATE OR REPLACE FUNCTION children_ids(id INTEGER)
RETURNS INT[]
LANGUAGE plpgsql
AS
$$
DECLARE
    ids INT[];
    r RECORD;
BEGIN
    FOR r IN WITH RECURSIVE t
    AS
    (
        SELECT * 
          FROM relations sa
         WHERE sa.id = id
         UNION ALL
        SELECT next.*
          FROM t prev
          JOIN relations next ON (next.parent_id = prev.id)
    )
    SELECT t.id FROM t
    LOOP
        ids := ids || r.id;
    END LOOP;

    RETURN ids;
END
$$;

然后，我需要查询另一个表，以获取函数返回的任何子id中的记录数：
我要执行的查询是：

SELECT COUNT(p.*) FROM books AS p WHERE p.id = ANY(children_ids(20));

但由于超时，它失败了。奇怪的是，在我的localhost中，这个相同的查询工作正常，数据库被克隆，PostgreSQL版本是相同的。
我已经测试了另一种变体，结果如下：

SELECT children_ids(20);
-- WORKS: {20}

SELECT COUNT(p.*) FROM books AS p WHERE p.id = ANY(children_ids(20)); 
-- FAILS: TIMEOUT

SELECT COUNT(*) FROM books AS p WHERE p.id = ANY(children_ids(20)); 
-- FAILS: TIMEOUT

SELECT * FROM books AS p WHERE p.id = ANY(children_ids(20));
-- FAILS: TIMEOUT

SELECT COUNT(p.*) FROM books AS p WHERE p.id = ANY(ARRAY[20]);
-- WORKS: 0

SELECT COUNT(p.*) FROM books AS p WHERE p.id IN (20);
-- WORKS: 0

SELECT COUNT(*) FROM books AS p WHERE p.id IN (20);
-- WORKS: 0

你知道这里会发生什么吗