Between 22 May at 22:45 UTC and 27 May at 11:35 UTC, Instagram publishing experienced significant degradation due to a regression in Meta's Instagram Graph API. The issue affected approximately 12,579 organizations, resulting in roughly 37,000 failed posts (17,000 above baseline), with single-image posts hit hardest at a 45% failure rate. Our engineering team implemented a temporary workaround on 25 May at 22:00 UTC by adjusting the token used for container verification, which restored success rates to 96.5%+ and allowed us to proactively notify 4,660 affected customers. Meta confirmed their fix on 27 May at 11:35 UTC, and we rolled back our workaround on 28 May. No Buffer-side issues contributed to the incident.

To prevent similar incidents in the future, we're making several improvements: adding faster detection alerts so we're notified within minutes rather than hours; logging Meta Graph API usage headers and response details to quickly confirm or rule out throttling versus upstream issues; refreshing our Meta support escalation runbook with current contacts and backup procedures; and documenting the Facebook Login Instagram token types (User vs Page) and making that information visible in logs and incident context so responders can spot divergent token behavior early. These changes will help us identify and respond to third-party API regressions more quickly and effectively.