Local Ablation or Lesion Probe¶

Lesion diagnostic — instantiates Neighborhood-Preserving Substrate Mapping

Deliberately disables one substrate region and measures exactly what degrades, so the real blast radius of a local failure — and whether it stays local — is known before it happens for real.

Local Ablation or Lesion Probe is a destructive test that switches off, corrupts, or removes a chosen substrate region and observes the resulting deficit. Its defining move is being active and deliberate: where a canary passively waits for a region to fail in production, an ablation probe causes the failure on purpose, under controlled conditions, to trace precisely which functions degrade and how far the damage spreads. It converts "we assume failures stay local" into a measured footprint — and in doing so it is the mechanism that produces the failure-footprint model the rest of the archetype's monitoring relies on.

Example¶

A ride-hailing platform maps cities onto geographic shards and claims that a shard outage only affects that one city. To check the claim rather than trust it, the reliability team runs a game-day: during a controlled window they deliberately kill one shard — the ablation — and watch everything with the rest of the system instrumented. Most of the blast radius is where expected: that city's driver-rider matching stops. But the probe reveals a surprise — a shared surge-pricing service silently fanned the failure out, briefly degrading two neighboring cities. That non-local footprint, discovered by lesion on purpose rather than by outage on a bad night, is exactly what the team needed: they add a bulkhead so the next real shard failure stays inside its region. The probe destroys deliberately in order to learn the shape of damage.

How it works¶

Pick a target region and a failure mode — clean removal, corruption, or slow degradation. Run it under controlled load with everything else instrumented, then measure which functions degrade and how far the effect reaches (the blast radius). Compare the observed footprint against the "should be local" expectation, and record the footprint as a reusable model. Two things set it apart: it injects the fault (a canary never does) and it runs offline and systematically to map a footprint in depth, where a broad perturbation sweep trades depth for breadth.

Tuning parameters¶

Lesion target and granularity — which region and how large a piece is disabled; fine lesions localize cause, coarse ones test worst cases.
Failure mode — clean removal versus corruption versus gradual degradation; each produces a different footprint.
Blast-radius scope — how wide you instrument to catch nonlocal effects; instrument too narrowly and fan-out is missed entirely.
Load conditions — quiescent versus peak; damage often spreads only under load, so a calm-hours lesion can look reassuringly local and lie.
Recovery observation — whether you also measure how the region heals once the lesion is lifted.

When it helps, and when it misleads¶

Its strength is converting an assumption of localized failure into evidence, and surfacing the hidden couplings that make damage nonlocal before a real incident does. Its limits are inherent to a destructive test. It is destructive, so it must run in a safe environment or a blast-radius-bounded slice — and it tests only the failure modes you thought to try, so an unimagined mode goes unprobed. The classic misuse is running a lesion in live production without bounding the blast radius and thereby causing the very outage you were trying to study. The discipline is to bound the blast radius, run in a controlled window, and feed the resulting footprint to the canaries that will watch for it in the wild.^[1]

How it implements the components¶

It owns the lesion protocol and produces the failure-footprint model; it does not monitor live or build the map:

perturbation_and_lesion_test_plan — it is the lesion protocol: what to disable, under what conditions, and what to measure.
local_failure_footprint_model — its output is the measured footprint of a local failure, the model other mechanisms then consume.

It does not passively watch for that footprint in production — that is the Canary Region Probe, which consumes this probe's output — and it is not the broad systematic sweep of the Perturbation-Response Map.

Instantiates: Neighborhood-Preserving Substrate Mapping — it tests the archetype's promise that local damage produces a bounded, understandable deficit.
Sibling mechanisms: Canary Region Probe · Perturbation-Response Map · Adaptive Remeshing and Reallocation · Adjacency-Matrix Preservation Test · Boundary and Seam Regression Test · Calibration Anchor Stimuli · Coverage-Hole Heatmap · Elastic-Net Embedding · Lateral-Interaction Smoothing · Magnification Function · Map Fold and Collision Scan · Map Registration and Alignment · Multidimensional Scaling Layout · Neighborhood Trustworthiness and Continuity Metric · Receptive-Field Tiling Grid · Self-Organizing Map Training · Topographic Error Measure · Versioned Coordinate Atlas

References¶

[1] Chaos engineering — deliberately injecting controlled failures to learn a system's real failure behavior and "blast radius" before an uncontrolled incident does. Cited as the real, correctly-used anchor. ↩