Report accepted Boolean retries on features and fuzz runs (#249)

Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

docs/robustness.md

@@ -36,6 +36,30 @@ rejected before any jobs or output directories are created. The maximum seed
 - `cargo test` runs a fixed set of seeds that must stay clean
   (`crates/engine/src/tests/robustness.rs`), next to the minimised cases fixed so far.
 
+## Successful Boolean retries
+
+When an accepted Boolean result needs a numerical retry, the feature shows a
+warning in the timeline. `timeline.inspect` and `doc.inspect` expose its
+`boolean_diagnostics`, including the strategy and the applied translation,
+rotation, face offset and tessellation tolerance in millimetres/radians.
+
+A common shift of both operands is reversed on the result. Coincident-face
+separation and bore offsets describe temporary working geometry; they are not
+claims about the final model's error. A tool nudge changes the relative operand
+placement and reports both its translation magnitude and rotation angle.
+Exact planar and piecewise fallbacks report zero displacement. The existing
+volume and spatial acceptance checks still apply to all results.
+
+Reports belong to the evaluated feature and survive cached rebuilds and
+undo/redo. Cold reopening regenerates them from the embedded design. Failed or
+discarded retry branches do not contribute successful-retry events. Reports
+are capped at 1024 events per feature and explicitly identify truncation.
+
+Fuzz JSON and the text report count successful retry events by strategy for
+designs and Boolean identity checks separately, including nested operations.
+These are event counts, so a single Boolean or seed can contribute several.
+Truncated reports are counted separately rather than implying complete totals.
+
 ## Results
 
 1000 seeds (0–999), 8 processes on a heavily loaded machine (load average 60–80), 90 s per