This issue is resolved. From 08:09 UTC on 13 October, a single malformed record blocked one of the queues we use to process thread events, which delayed a small share of thread updates across workspaces.

Messages were received as normal throughout. We've deployed a fix and processed the delayed events, and everything is back to normal. If a thread still looks out of place, please get in touch and we'll take a look.

We've deployed a fix, and delayed thread events are now being processed. Affected threads are catching up, so you may still see some catch-up activity for a short while.

Messages were received as normal throughout. We're monitoring closely and will post another update once everything has fully caught up.

We've identified the cause. A single malformed record was blocking one of the queues we use to process thread events, which affects a small share of thread events (around 2%) across workspaces. Messages are still being received as normal. We're preparing to clear the backlog shortly, and when we do you may see a short burst of catch-up activity.

We'll post another update once the backlog has been processed.

We're continuing to investigate delayed processing of some thread events. Affected threads may not update as expected. For example, a thread may not be assigned or reopened, an SLA timer may not stop when a reply is sent, or a snoozed thread may not move back to needing a response when a customer replies. Messages are still being received. We've found the part of our system that's blocked and are carefully working on a recovery, so that catching up doesn't trigger a burst of delayed actions. We'll share another update as soon as we have more information.