Etsy recommend better incident retrospectives with a focus on learning from the incidents. Before the retro review the timeline to prepare initial questions. During the retro, emphasize asking effective questions. Defer focus on remediation for a few days. pdf 
Reliability is not the absence of incidents. Rather, reliability emerges from people's expertise.
Etsy conducts an internal course to build a corps of effective facilitators. Training includes: 3 seminars covering theory, empirical foundations, case studies, readings, and written exercises; an interactive workshop to practice core skills; shadowing for new facilitators; regular follow-up discussion groups. The text book for the course is The Field Guide to Understanding Human Error, by Sidney Dekker. book
Facilitators are not looking for causes. The debriefing exists to hear people's stories of what they did. Incidents provide unique opportunities to examine the actual work required to keep systems running (as opposed to what we imagine that work to be).
Tell multiple stories. Resist the temptation to reduce it to a single narrative. Complexity of the event, not simplicity, is what we need to understand. There are always many points of view, none omniscient. Participants in the retro benefit from hearing the varied, overlapping, and contrasting perspectives. Each observer will construct their own understanding of what happened, informed by their expertise and place in the system.
Everyone is an expert.
When we ask for descriptions of their work, people share their expertise. We hear tips and tricks from a full career: shell aliases, bookmarks, one-liner scripts, the trustworthy charts for diagnosing specific problems. Which procedures do they always follow? Which procedures do they avoid & why? We learn how their usual response to that alert leads to recovery. We learn under what conditions they deviate and why. We coax out the hidden nuances underlying their actions, decisions, and rationales.
Discover what makes people’s work difficult, and what makes them good at it, despite the difficulty.
Vicarious Experience
Rich and complex stories are compelling. They stick with an audience. They accelerate and augment training and shared expertise.
# Preparation
Gather in a room together (video conference for remote participants). Make the initial learning dynamic and conversational, not static.
Understand the timeline ahead of time. Identify the boundaries to focus the conversation. What were the most interesting communication patterns, hypotheses, decisions, observations, and actions? There will never be enough time to cover everything. Review the timeline to plan the retro agenda and budget time to ensure you can get to the richest opportunities for learning.
Figure out who needs to join the retro. Remember to include the diverse perspectives to ensure diverse learning.
It’s crucial to establish trust with all participants up front. Gauge how much fear or anxiety they’re feeling. Tread with sensitivity here to set everyone up for success in the retro. Help them feel at ease. Remind them that feeling anxious or embarrassed is what makes them the best teachers for their coworkers, who will find themselves in similar situations in the future. Remind them hindsight and after-the-fact knowledge of the outcome is what makes it embarrassing. Remind them that before the event, they were just doing normal work on a normal day.
Note the important pivot points in the timeline to identify questions likely to draw out the rich context of those moments in the event.
Prepare to ask for subjective data – opinions, judgements , assumptions, beliefs, theories – to create context about actions and decisions taken.
One of the hardest parts is making everybody understand that we are not coming together to prevent a future event from happening. This event may very well happen again. Preventing the future is not the goal for the retro — learning is. To that end, learning needs to be the entire focus. Trust that everyone will naturally make efforts at prevention — let it be a side-effect of effective learning.
# Agenda
For a 60 min meeting, spend 35-40 min reconstructing the timeline. Get a reconstruction that everyone can agree is as close to possible what happened. When people suggest remediation, ask them to write it down and save it for later. Remain focused on filling in the timeline from diverse perspectives. Ask what's missing or if anyone has left anything out.
For the next 10-15 min, brainstorm what was learned. Invite wide ranging, even haphazard ideas and continue to resist actionable tickets. In the last 10 min, collect the suggestions, select a small group of people from the meeting and ask them to reflect on the brainstorm list. Let the suggestions marinate.
In a couple days or so, bring the group back together to discuss remediation. Generally, insights appear when we aren't forcing them and when we're allowed time to gain some perspective. Give time to ensure tasks are effective, logical and stand the test of time, rather than racing to the first idea. Often, hasty intervention can needlessly complicate and introduce more new failure modes than the things they aimed to prevent.
# The Art of Asking Questions
Ask "How?" instead of asking "Why?" in order to steer discussion toward the humble nuts and bolts of what actually happened. "Why?" leads toward cognitive landmines and traps.
What was observed?
Uncover the context of assessments or judgements.
What rationale led to specific decisions?
What is known? Is it assumed to be common knowledge?
What was their state of mind?
Explore mental models for how things "should" work.
What factors led to specific actions?
What signals brought people to ask for help? How did they know whom to ask?
We were surprised.
When did you first notice this didn't look right?
Were you confident about what's happening at this point?
Something was different this time.
What made that metric look suspicious?
Anything special others should know about that error log?
What indicators do you expect for this type of failure?
Complexity was unfolding.
How do you think about which changes to make?
What other options were you considering?
Can you draw what you thought it was doing?
How do you decide if the situation is getting worse?
When in doubt, ask dumb questions. Let people showcase their expertise, or surface their tacit knowledge.
Expect to build skill in facilitation. The art includes knowing when to dig into the details, when to shift to the big picture, when to compare alternate perspectives. If a person claims a thing is obvious, validate that belief with the rest of the room.
How much of this is new information for everyone?
How many of you knew all of the moving pieces here?