1. It's too configurable. It's like the company wiki. People come in, set up some random projects, processes, fields, statuses, etc. and then move on. There's rarely someone making it all consistent and sensible. This results in task statuses that can be TODO, BACKLOG, PENDING, WAITING, TO DO, etc. etc. etc. You also end up with waaaay too many fields for tasks. Arbitrary distinctions between "Tasks" and "Stories", etc. You're at the mercy of your Jira admin who will definitely make worse decisions than e.g. the developers of Phabricator or Gitlab. Hiding useful fields, adding pointless ones, etc.

2. Despite being super configurable, it can't do some really basic things you'd expect from something whose sole job is task tracking. For example you can only parent tasks 2 or 3 levels deep. A task can't have two parents. Subtasks can't be in different sprints. You can't reorder tasks by priority in the backlog.

3. Most offensively it is just incredibly slow. The main way you add tasks to a sprint is drag and drop on the backlog, but it performs so badly (100% CPU all the time) that they've had to add context menu options to move tasks to the top/bottom or to a sprint. I believe there's also an option to disable animations. A simple web page should not consume 100% of my CPU.

It's by far the worst issue tracker I've ever used.

Companies still pay for it though because PMs love the pretty burndown charts and being able to add a gazillion fields to tasks. How will they report project status to their bosses if they can't have Jira work out the exact-to-the-second estimated delivery time automatically? And by "automatically" I mean by making someone else do all the work.