Hi TG,
I completely understand your frustration with these persistent Copilot crashes - this is incredibly disruptive when you're trying to maintain productivity in your workflow. You've already done solid troubleshooting work, but let me walk you through some additional approaches that often resolve these issues.
Session Limits and Resource Management
The pattern you're describing (5 interactions before slowdown, crashes at 15-20 questions) strongly suggests you're hitting resource or session limits. Microsoft 365 Copilot has built-in conversation limits and memory constraints that aren't always clearly communicated:
- Conversation Length: Copilot sessions have practical limits around 20-30 exchanges before performance degrades
- Memory Usage: Each interaction builds up context that consumes resources
- Recommendation: Start fresh conversations every 10-15 interactions as a preventive measure
Clean Environment Testing
Let's isolate whether this is environment-specific:
- Create a new user profile in Windows and test Copilot there
- Try in Private/Incognito mode across all browsers
- Test during off-peak hours (early morning/late evening) to rule out server load issues
- Disable all browser extensions temporarily, especially ad blockers and productivity tools
Resource Monitoring
The error code you mentioned (-536870904) often indicates memory issues. Here's how to monitor:
- Open Task Manager while using Copilot
- Watch memory usage of your browser/app - look for steady increases
- Monitor CPU spikes during interactions
- Check available disk space - ensure you have at least 2GB free
- Close other applications while using Copilot to free up resources
Enhanced Cache Reset Process
You mentioned clearing cache, but let's be more thorough:
For Browsers:
- Clear browsing data for "All time"
- Include: Cookies, cached files, site data, hosted app data
- Edge specifically: Go to
edge://settings/reset and reset settings
- Chrome: Clear "Cookies and other site data" and "Cached images and files"
For Microsoft 365 Desktop App:
- Press
Win+R, type %appdata%\Microsoft\Microsoft 365
- Delete the "Copilot" folder if it exists
- Restart the app
Working Within Interaction Limits
Best Practices:
- Break complex tasks into smaller conversations
- Save important responses externally before continuing long sessions
- Use specific, focused questions rather than building on long context
- Consider starting new conversations for different topics/projects
Service Health Monitoring
Before diving into deep troubleshooting:
- Check Microsoft 365 Service Health:
admin.microsoft.com > Health > Service health
- Monitor @MSFT365Status on Twitter/X for real-time updates
- Track your crash times - if they correlate with service issues, that's your answer
Escalation with Proper Logging
Since your IT team couldn't help, let's gather the right data for Microsoft support:
Collect These Logs:
- Browser Developer Tools: Press F12, go to Console tab, capture any red errors during crashes
- Windows Event Logs:
- Open Event Viewer
- Navigate to Windows Logs > Application
- Look for errors around crash times
- Network Activity: In Developer Tools, go to Network tab and capture failed requests
- Session Details: Note exact time of crashes, number of interactions, type of questions asked
For Escalation:
- Document the pattern clearly (X interactions before slowdown, Y before crash)
- Include your tenant ID if known
- Provide browser/app version numbers
- Include the error codes you captured
While working on the permanent fix:
- Set a timer for every 10 interactions to remind yourself to start a new conversation
- Save key information in a separate document as you go
- Use multiple browser tabs/windows to maintain separate Copilot sessions for different tasks
- Consider mobile app as a backup when desktop crashes
Next Steps
If the clean environment testing still shows crashes, this is likely a backend issue that needs Microsoft engineering attention. The fact that multiple users are experiencing this (as evidenced by the closed thread you referenced) suggests it's not isolated to your setup.
Try the resource monitoring and clean environment testing first, then escalate with the detailed logs if the issue persists.
Let me know what you discover with these steps - I'm happy to help you interpret the results and guide the next phase of troubleshooting!
Best regards