Fewer Mystery Errors When the API Switches Over During Peak Nights
Conference night is the worst possible moment for a "something went wrong" error that nobody can explain. In some cases, these errors weren't caused by anything the user did—they happened when the backend routed a request to a new API target mid-session during a failover event.
What was causing the errors
During periods of high traffic—exactly the moments when many families are booking simultaneously—the backend load balancer occasionally switched a session's request routing mid-flight. The receiving endpoint didn't have the session context it expected, and the result was an opaque error message with no actionable guidance for the user.
How it's fixed
We fixed the session routing logic to maintain context continuity across failover transitions. Sessions that were mid-request during a routing switch now complete successfully rather than failing. The fix also improves the error message in the rare cases where a routing event does affect a session—families see a specific message asking them to refresh, rather than a generic system error.
For schools experiencing high concurrent booking traffic, this improvement pairs with the in-place agenda update improvements to create a more stable booking experience across the entire conference window.