Jobs are waiting in the queue longer than usual because capacity is tight. Queued jobs are safe: they keep their place in line and start as capacity frees up, and a job that cannot run by its deadline expires without being charged for compute it never received.
Usage and billing data has caught up: your usage page is current again, and invoices, receipts, and credits are landing normally. If something you expected from this window is still missing, file a ticket and we will reconcile it.
Usage and billing data is updating more slowly than usual. Recent activity may not appear on your usage page yet, and a new invoice, receipt, or credit may take longer than usual to show up. We are working on it now. Your nodes and jobs are not affected and keep running normally, and nothing is lost: the underlying usage, payments, and credits are all on record, and your pages fill in as we clear the backlog.
Sandbox command reliability is back to normal. Voided work was never billed; re-run anything you are missing on a fresh fork.
An elevated share of commands running in sandboxes is being cut short by infrastructure failures on our side. Any result marked void was not billed and was excluded from its batch - it is a gap in your data, not a wrong answer - and retrying it on a fresh fork is safe. We are investigating the cause. Nodes and jobs run on separate machines and are not affected.
The job queue is draining normally again.
Job interruptions are back to normal levels.
Some running jobs are being interrupted and returned to the queue more often than we consider acceptable. Interrupted jobs retry automatically - you do not need to resubmit - but affected jobs will take longer to finish than usual.
Command status is up to date again.
The status of commands run on nodes is updating more slowly than usual: a command may show as still running after it has actually finished. The commands themselves and their output are not affected.