Queries like this are pretty common in our slow query logs:

228826-# User@Host: fetlife[fetlife] @ app4.dal.fetlife [10.18.5.79]
228827-# Thread_id: 42698 Schema: fetlife_production
228828-# Query_time: 7.656149 Lock_time: 0.000023 Rows_sent: 0 Rows_examined: 0 Rows_affected: 1 Rows_read: 2
228829:INSERT INTO `friendship_requests` (`user_id`, `requested_by_user_id`, `created_at`, `updated_at`) VALUES (?, ?, '2012-02-21 03:37:50', '2012-02-21 03:37:50');

This table has been behaving funny for a while. Almost no lock time, only 2 rows read, 1 affected, and almost 8 seconds of runtime.

CREATE TABLE `friendship_requests` (
`id` int(11) NOT NULL auto_increment,
`user_id` int(11) NOT NULL,
`requested_by_user_id` int(11) NOT NULL,
`created_at` datetime default NULL,
`updated_at` datetime default NULL,
PRIMARY KEY (`id`),
UNIQUE KEY `index_friendship_requests_on_user_id_and_requeste d_by_user_ id` (`user_id`,`requested_by_user_id`),
KEY `index_friendship_requests_on_user_id_and_created_ at` (`user_id`,`created_at`)
) ENGINE=InnoDB AUTO_INCREMENT=13940603 DEFAULT CHARSET=utf8

and

+----------+
| count(*) |
+----------+
| 983848 |
+----------+

No other table in our system has this issue.

Ideas?