The changes described in our previous update have held up, and sandbox connections have been working normally since then. We are marking this incident as resolved.

We sincerely apologize for the disruption to your sandboxes. We are working on further improvements to prevent this from happening again.

We have deployed a fix for the internal DNS issue that caused sandbox tabs to show a "Connecting" or loading page. Since 07:03 UTC, connections to sandboxes are succeeding normally, and DNS response times are back below normal levels.

The issue was caused by a sharp rise in DNS lookups from a large number of sandboxes starting at the same time. That load exceeded a capacity limit in our internal DNS layer, so a portion of lookups failed or were delayed. Participants then saw tabs that loaded slowly, reconnected repeatedly, or dropped back to the loading page. We have removed that limit and increased the capacity of the DNS service so it can absorb this kind of load.

Sandboxes kept running throughout and no lab progress was lost. If a sandbox tab is still showing the loading page, refreshing it should now connect right away.

We are continuing to monitor and will post a final update once we are confident the issue is fully resolved.

Since approximately 02:50 UTC today, some participants are seeing a "Connecting"
or loading page instead of their sandbox tabs (terminals, web services, IDEs).
The underlying cause is an intermittent failure in the internal DNS layer that
routes traffic from the platform to sandbox hosts. Sandboxes themselves remain
running and are not affected; the problem is in reaching them.

Symptoms you may see:
- A loading page that reloads repeatedly or takes much longer than usual to connect
- A tab that connects, then drops back to the loading page
- Setup or lifecycle scripts taking longer than usual to complete

Sandbox creation and cleanup are working. Refreshing the tab usually succeeds
within a few attempts, as resolution is failing intermittently rather than
completely.

Our engineering team has identified the affected component and is actively
working on a fix. We will post an update here as soon as we have more
information or the issue is resolved.