cyb50-quant/research/dragon/v2/openspec/changes/glued-followthrough-shadow-reentry/proposal.md

Why

buy_block_glued_high_weak_rebound currently acts as a hard veto with no structured "blocked now, re-confirm later" path. This is good for RC1 stability, but it also means we cannot observe delayed followthrough behavior in a controlled, testable way.

Recent review suggests the correct direction is not to relax RC1 hard-veto defaults. The correct direction is to add a shadow-only research scaffold that can evaluate delayed reentry after explicit confirmation.

What Changes

• Add a dedicated glued followthrough confirmation module.
• Add branch-gated pending lifecycle for blocked glued weak-rebound signals.
• Add a shadow research branch: alpha_first_glued_followthrough_probe.
• Keep RC1 default branch behavior unchanged.
• Add regression tests:
• default local baseline hashes unchanged
• probe branch fixture behavior pinned
• Wire shadow branch into daily signal pipeline output only.

Capabilities

New Capabilities

• glued-followthrough-shadow-pending: subtype-aware pending and confirmation evaluation for blocked glued weak rebounds.
• shadow-branch-observation: expose probe branch in daily pipeline without changing forward default governance path.

Modified Capabilities

• dragon-strategy-compatibility-facade: supports optional glued pending lifecycle in branch-gated mode.

Impact

• Affected code: dragon_strategy.py, dragon_strategy_config.py, dragon_branch_configs.py, dragon_rule_catalog.py, dragon_daily_signal_pipeline.py, new glued followthrough module, new tests.
• Runtime impact: negligible.
• Governance impact: none to RC1 default branch; this is shadow research only.