LLM crawlers DDoS SourceHut's git hosting into 168 hours of tracked downtime
SourceHut · Mar 17, 2025 · General agent
What happened
SourceHut's public status page tracked a git.sr.ht disruption caused by what the maintainers described as aggressive LLM crawlers continuing to DDoS the service, resolved after 168 hours 30 minutes of tracked downtime. The team's fix was deploying Anubis, a proof-of-work challenge proxy that makes uncontrolled automated clients pay compute per request. The Register reported the git-hosting service had faced repeated disruptions from crawlers making excessive data demands.
Causal vector
High-volume automated clients from many IPs overwhelming origin infrastructure; no effective throttling layer in front of the service
Source
Reported by SourceHut status page (git.sr.ht). Verified against the primary report.
Unthrottled autonomous clients are the inbound twin of runaway spend. The velocity limit that stops a looping agent from draining a budget is the same control SourceHut rebuilt as Anubis: a hard rate barrier that makes uncontrolled automation too expensive to continue.
The six rule types that contain this class of failure
Per-transaction cap
Any single spend above your ceiling is BLOCKED before it moves.
Daily total
Cumulative spend across all agent calls, bounded per day.
Velocity limit
Stops runaway retry loops — the #1 cause of overnight losses.
Merchant allowlist
Only approved destinations can ever receive funds.
Category rules
Flag high-risk classes (crypto, infra, refunds) for review.
Approval threshold
Above a value, the action waits for a human.
Related incidents
Microsoft EVP: 'tokenmaxxing is not what we are optimizing for' - division-level AI token budgets imposed as engineers spend thousands each per month
Microsoft · Aug 4, 2026
Three-person agency hit with a $14,000 AWS Bedrock bill after attackers burn stolen keys on Claude
Amazon Bedrock (customer: three-person agency) · Jul 16, 2026
AWS user faces a $30K–$38K Bedrock bill that cost-anomaly detection never flagged
AWS Bedrock customer (individual developer) · May 14, 2026
Don't be the next entry
Every incident in this database is the result of trusting a prompt, a provider cap, or a human review cycle. sipi.bot replaces all three with one deterministic call. 90 documented failures, one control.