Hold it soon, keep it short
Meet within a week, while details are fresh, and keep it to thirty minutes. Invite the agents who handled the tickets and one person from engineering.
Build the timeline from tickets
List when the first ticket arrived, when you noticed, when you told customers, and when it was fixed. The ticket timestamps give a more honest view than memory.
Look at what customers saw
Read a sample of tickets. What did people try? What did they expect? Note any wording in your status messages that confused them.
Keep it blameless
Ask what made the slow step easy to take, not who took it. Write down gaps in tools, templates and handoffs rather than names.
Follow up with the customers
Send a short note to everyone affected once the fix is confirmed: what happened, what is fixed, and what you are changing. Customers remember the follow-up more than the incident.
Change something concrete
End with two or three actions, each with an owner and a date: update a canned reply, add a status page step, change routing for the next incident. Check them at the next weekly review.