Fix PR #48 review finding: Doppler horizon compounding

[daedalus]

Addresses the Copilot thread on #48.

Bug (core/power_doppler.py)

• Each returning dropped seed doubled one shared horizon: N returns → ×2^N, abandonment effectively off, gone seeds pin every slot.
• Also found: dropped-key memory was max_seeds keys, so with large corpora keys were forgotten before returning — horizon never widened, zero frames closed (200 seeds / 4 slots → 0 ensembles).

Fix

• Remember each dropped key's last-touch tick (LRU, max_dropped = 2^15, ~5 MiB).
• On return: horizon = max(horizon, 2 × gap). Tracks the slowest real revisit gap; never compounds.

Tests

• test_regression_many_returns_do_not_compound_horizon (falsification, failing before fix): 200 seeds / 4 slots → horizon ≤ 4×cycle and frames close (now 400, 32 ensembles; before 0).
• test_adversarial_abandoned_keys_do_not_grow_horizon: non-returning keys leave horizon unchanged; memory bounded by max_dropped.
• 98 passed (test_power_doppler.py, test_schedules.py); ruff + lizard clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Rvj2gdMAa4HA8MGHULfDJE

---

Generated by Claude Code

Summary by Sourcery

Fix Doppler horizon management so slow returning seeds are accommodated without disabling frame abandonment or allowing unbounded state growth.

Bug Fixes:

• Prevent Doppler horizon growth from compounding across returning seeds by basing it on the largest measured revisit gap.
• Preserve dropped-seed revisit history with a bounded LRU so slow corpus cycles can widen the horizon correctly instead of losing keys prematurely.

Documentation:

• Update Doppler schedule documentation and changelog to describe bounded dropped-key tracking and non-compounding horizon behavior.

Tests:

• Add regression coverage for large corpus cycles, abandoned keys, and invalid dropped-key capacity configuration.

[sourcery-ai]

Reviewer's Guide

Fixes Doppler horizon runaway by tracking dropped seeds’ last-touch ticks and setting the horizon from the slowest observed revisit gap, while expanding bounded key retention so large corpus cycles are measured; adds regression coverage and updates documentation.

Sequence diagram for Doppler dropped-seed revisit handling

sequenceDiagram
    participant Scheduler
    participant PowerDoppler
    participant DroppedKeys
    participant OpenFrames

    Scheduler->>PowerDoppler: process seed key
    PowerDoppler->>DroppedKeys: pop(seed_key)
    alt key was previously dropped
        PowerDoppler->>PowerDoppler: compute gap = ticks - touched
        PowerDoppler->>PowerDoppler: set _stale_after = max(_stale_after, 2 * gap)
    end
    PowerDoppler->>OpenFrames: _admit()
    alt frame is abandoned
        OpenFrames-->>PowerDoppler: old frame and old.touched
        PowerDoppler->>DroppedKeys: store seed key -> old.touched
    end

Loading

Flow diagram for non-compounding Doppler horizon

flowchart LR
    A[Seed frame is dropped] --> B[Remember key and last-touch tick]
    B --> C[Seed returns after measured gap]
    C --> D[Compute 2 × gap]
    D --> E[Horizon = max current horizon, 2 × gap]
    E --> F[Slowest observed revisit gap controls abandonment]

Loading

File-Level Changes

Change	Details	Files
Replace compounding horizon growth with revisit-gap-based widening and retain sufficient dropped-key history.	Store each dropped frame’s last-touch tick in a bounded LRU cache. On return, widen the stale horizon to at least twice the measured gap instead of multiplying it per return. Expose and validate the configurable dropped-key cache capacity, defaulting to 2¹⁵ entries.	src/fuzzer_tool/core/power_doppler.py
Add regression and boundary tests covering horizon stability, abandoned keys, and configuration validation.	Exercise a 200-seed/4-slot cycle to verify frames close and the horizon remains bounded. Verify abandoned keys do not change the horizon and dropped-key memory respects its configured limit. Verify invalid dropped-key capacity is rejected.	tests/test_power_doppler.py
Document the corrected Doppler horizon behavior and dropped-key memory bounds.	Update the changelog with the compounding and forgotten-key fixes. Update the Doppler deep-dive documentation to describe measured-gap widening and the bounded dropped-key LRU.	CHANGELOG.md docs/DEEP_DIVE.md

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sourcery-ai]

Sorry @daedalus, you've used your own review budget of 250,000 diff characters for the last 7 days.

You can request another review in 1 day and 13 hours by commenting @sourcery-ai review. Upgrade to get a review now.

[Copilot]

Copilot review overview

🟡 Changes recommended

The fixed LRU capacity preserves the zero-frame failure for unlimited corpora exceeding 32,768 seeds.

Review effort: Balanced
Findings: 1

Open (1)

• Fixed-size LRU misses cyclic revisits when corpus exceeds capacity · New

What changed in this PR

Fixes Doppler horizon compounding and adds bounded dropped-seed tracking.

Changes:

• Derives horizon growth from measured revisit gaps.
• Adds bounded tracking and regression tests.
• Updates schedule documentation and changelog.

File	Description
src/​fuzzer_tool/​core/​power_doppler.py	Revises horizon adaptation and dropped-key retention.
tests/​test_power_doppler.py	Tests compounding, bounds, and validation.
docs/​DEEP_DIVE.md	Documents revised behavior.
CHANGELOG.md	Records the fix.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.