How is this different from the static blast-radius constitution in Claim 55?

Claim 55 assigns a fixed blast-radius tier to an action type before it ever runs. This mechanism watches for a real-time pattern - the same fault signature matching across many sites at once - and dynamically upgrades the effective classification of that entire pattern at decision time, even though each individual occurrence might statically classify as low-impact.

Why suppress remediation instead of just letting each site fix itself independently?

A fault appearing simultaneously across a large share of the fleet is far more likely to share a common root cause - a bad deploy, a shared dependency failure, a config push - than to be many unrelated coincidences. Letting every site independently and autonomously apply its own contained fix in that situation risks masking the real root cause or, worse, having many uncoordinated autonomous actions interact badly with each other.

What are the two conditions that trigger reclassification?

Both a minimum number of distinct sites where the same fault signature has matched, and a minimum proportion of the overall monitored fleet, must be crossed within a common time window before the reclassification and suppression take effect. Either condition alone is not sufficient.

Correlation-Triggered Blast-Radius Reclassification

Executive Summary

Claim 55 classifies a candidate remediation's blast radius statically, per action type, before it is ever attempted. That static classification has a blind spot: an action that is genuinely low-impact when it happens once, in isolation, can be part of something much larger when it is happening everywhere at once. AEGIS closes that blind spot with a correlation detector that watches, in real time, how many distinct sites and what share of the overall fleet are matching the same fault signature within a common time window.

When both a minimum-site count and a minimum fleet-proportion threshold are crossed simultaneously, the system reclassifies the entire matching pattern - occurrences that would individually be assessed as contained or cell-scoped - into a single fleet-scoped blast-radius classification, and suppresses autonomous remediation across all of them in favor of an escalated response. This happens regardless of how confident any individual occurrence's fix would otherwise be under Claim 54.

Core Innovation: Blast-radius classification is not fixed at decision time by action type alone; it can be dynamically upgraded in real time when cross-site correlation reveals that a pattern of individually-small actions is actually one systemic event, suppressing what would otherwise be many independently-authorized autonomous fixes.

Problem Statement

A blast-radius model that only looks at one action in isolation cannot see the pattern formed by many simultaneous occurrences of the same fault. A restart, a cache flush, or a config rollback that is genuinely safe to run autonomously on one node can, if triggered simultaneously and independently across a large share of a fleet by the same underlying systemic cause, produce exactly the kind of coordinated, cascading impact that a static per-action classification was never designed to catch.

70%

of outages Uptime Institute's annual survey attributes at least partly to human error or process failure

$9,000

widely-cited (Ponemon Institute) estimate of average cost per minute of unplanned downtime

34%

of the monitored fleet matching the same fault signature simultaneously - one of the two thresholds that triggers reclassification

Why Existing Solutions Are Insufficient

Static per-action blast-radius policies: Classify an action type once and apply that classification uniformly, with no mechanism to notice that the same action is currently happening everywhere at once.
Alert-storm suppression/deduplication (common in AIOps platforms): Groups and deduplicates alerts for human visibility, but typically does not itself gate whether underlying automated remediation actions are allowed to keep firing autonomously during the storm.
Rate limiting on individual automated actions: Caps how fast a single action type can fire, but a rate limit calibrated for normal operation is not the same as a correlation-aware circuit breaker that recognizes a specific cross-site pattern as systemic.
Manual "declare an incident" processes: Depend on a human noticing the cross-site pattern and manually pausing automation, introducing exactly the detection lag this mechanism is designed to remove.

Technical Architecture

Cross-Site Correlation Detection

The system continuously tracks, per fault signature, the set of distinct sites at which that signature has matched within a common, bounded time window, alongside what proportion of the total monitored fleet those sites represent.

Dual-Threshold Trigger

Reclassification triggers only when both a minimum number of distinct matching sites and a minimum proportion of the overall fleet are crossed within the same window. Either signal alone - many sites in a very large fleet, or a small number of sites that happen to be a large share of a small fleet - is evaluated jointly rather than in isolation, so the trigger reflects genuine breadth of impact rather than either raw count or raw proportion alone.

Reclassification and Suppression

Once triggered, every matching occurrence across the correlated pattern - regardless of its individually-computed blast-radius tier or confidence score under Claim 54 - is reclassified into a single fleet-scoped blast-radius classification, and autonomous execution is suppressed across the whole pattern in favor of an escalated response.

Correlation Detection & Suppression Logic

Condition	Description	Effect
Site-count threshold	Minimum number of distinct sites matching the same fault signature in-window	Necessary but not sufficient alone
Fleet-proportion threshold	Matching sites represent at least a defined share of the total monitored fleet	Necessary but not sufficient alone
Both crossed simultaneously	Combined signal of genuine systemic breadth	Triggers reclassification and suppression

This mechanism operates as a dynamic, decision-time refinement layered on top of Claim 55's static blast-radius baseline: the static tiers still apply to every individually-evaluated action, but this correlation detector can override the outcome for a specific pattern of simultaneous occurrences, upgrading them collectively to fleet scope even when none of them would individually classify that way.

Prior Art Differentiation

System	Cross-Site Correlation Aware?	Dynamic Reclassification?	Auto-Suppresses Remediation?
Static per-action blast-radius policy	No	No	No
Alert-storm deduplication (AIOps)	Yes, for alerts	No (visibility only)	Typically no
Per-action rate limiting	No	No	Partial (throughput cap only)
JIL Sovereign AEGIS	Yes, dual-threshold	Yes, real-time	Yes, fleet-wide

Patent Claim

Independent Claim 60: A computer-implemented method for dynamically reclassifying the blast radius of concurrent fault occurrences, comprising: detecting that a common fault signature has matched at or above a first threshold number of distinct sites and at or above a second threshold proportion of a monitored fleet within a common time window; reclassifying the plurality of matching fault occurrences, which would individually be classified as contained or cell-scoped, into a single fleet-scoped blast-radius classification; and suppressing autonomous execution of remediation actions across the reclassified occurrences in favor of an escalated response, notwithstanding any individual occurrence's independently-computed blast-radius classification or confidence score.