DevOps incidents jump 21% as downtime hits 9,255 hours

GitProtect has published research showing that incidents across major DevOps platforms rose 21% in 2025, while total disruption time climbed to 9,255 hours.

The study recorded 607 incidents in 2025, up from 502 a year earlier, and found that cumulative service disruption nearly doubled from 4,755 hours. It also identified 156 critical or major incidents, with a combined duration of 1,769 hours and 43 minutes, a 69% year-on-year increase in the most serious category.

The findings point to rising operational strain across tools used by software development and IT teams. Instability was most pronounced in the second quarter, when 47 incidents were logged, while June was the most volatile month, with 21 incidents.

Outage patterns

Degraded performance accounted for 62% of all recorded incidents, or 374 cases, making it the most common problem in the data. Even so, those events made up 34% of total downtime, or 3,059 hours, suggesting many were widespread but shorter than other types of incidents.

Maintenance made up only 4% of incidents but accounted for 30% of total disruption time, making it the biggest single driver of platform unavailability by duration despite its relatively low frequency.

Critical and major incidents represented 26% of all incidents and 22% of total disruption time. Minor issues with no direct operational impact accounted for 8% of incidents and 16% of total duration.

Platform breakdown

Among the platforms analysed, GitLab recorded the highest number of critical and major incidents in 2025, with 62 incidents and more than 754 hours of impact. Jira followed with 44 incidents and nearly 728 hours of downtime.

GitHub saw 35 critical and major incidents with 172 hours of impact. Bitbucket recorded 15 such incidents, totalling 113 hours and 59 minutes.

The report also highlighted sharp monthly swings in disruption levels across individual platforms. For GitLab, July was the worst month, with 11 incidents causing more than 252 hours of disruption, including a nearly 50-hour outage linked to incorrectly deleted OAuth refresh tokens.

GitHub recorded its highest monthly incident count in December, with five critical and major incidents. Its most severe monthly impact came in March, when two incidents caused more than 58 hours of degraded performance, including a prolonged outage triggered by expired internal credentials.

Bitbucket's most disruptive month was May, with five incidents producing more than 84 hours of impact, including a 49-hour disruption affecting pipeline execution. In Jira, June brought nine incidents and almost 48 hours of disruption. The longest single incident that month lasted almost 24 hours and affected Jira Software and Jira Service Management because of Forge-related errors in the Singapore region.

Operational cost

GitProtect attached an estimate of lost engineering productivity to the cumulative outage time. Using an average developer rate of USD $80 an hour, it estimated that the total disruption equated to more than USD $740,000 in lost productivity alone.

That calculation does not include wider commercial effects such as delayed releases, service credits, customer support costs or the impact on internal IT operations. The figures nevertheless underline how disruption in developer tools can directly affect engineering teams that rely on cloud-based services for code hosting, collaboration, deployment and ticketing.

The research covered security breaches, outages and performance degradation across major development platforms used by millions of people. Its figures suggest growing pressure on cloud-based software development infrastructure, not only in the number of incidents but also in how long teams remain affected when problems occur.

The data suggests that even widely used platforms with established resilience measures remain exposed to service interruptions. It also indicates that rising downtime, rather than incident count alone, is the clearest sign of a changing risk profile for development teams.

Maintenance windows, degraded performance and severe incidents each contributed differently to the overall disruption picture, with planned and unplanned maintenance accounting for a larger share of total downtime than the number of events alone might suggest. The pattern shows that frequency by itself does not capture the operational burden placed on users.

The analysis concludes that the gap between incident volume and incident impact has widened, as longer disruptions account for a growing share of overall downtime across the DevOps software market.