diff --git a/docs/session_summaries/2026-01-04-refactor-orbital-elements.md b/docs/session_summaries/2026-01-04-refactor-orbital-elements.md
new file mode 100644
index 0000000..592951c
--- /dev/null
+++ b/docs/session_summaries/2026-01-04-refactor-orbital-elements.md
@@ -0,0 +1,25 @@
+# Refactor Orbital Elements Session (2026-01-04)
+
+## Changes Made
+
+Refactored orbital elements calculation from test-specific to reusable code:
+- Moved `OrbitalElements` struct from test file to `src/bodies.h`
+- Moved `calculate_orbital_elements()` function to `src/bodies.cpp`
+- Created `tests/test_comet_orbit.cpp` with SOI transition tracking
+- Added `ParentChange` tracking to detect when comet switches gravitational parent
+- Removed `tests/test_soi_transitions.cpp` (not useful for detecting actual transitions)
+- Updated Makefile to include new test file
+
+## Commits
+
+- `0e2f9dd` Refactor orbital elements to reusable code and track SOI transitions
+
+## Results
+
+4 files changed, +225/-1 (net +224 lines)
+
+Orbital elements calculation is now reusable across the codebase. Test properly tracks parent index changes, though current configuration shows comet misses Mars's SOI by ~0.0004 AU (likely due to 60s time step granularity).
+
+## Outstanding Issue
+
+Test fails to reproduce manual debugging results. Manual debugging showed `parent_index` changing to 2 (Mars) at sim->time ≈ 133914000s (~1550 days), but automated test detects no parent changes. Closest approach in test: 0.004200 AU, Mars SOI: ~0.00379 AU. Discrepancy may be due to time step granularity (60s), different initial conditions, or timing of SOI calculations. Requires further investigation.