Follow up items for the week of July 18, 2026

[espg]

Summary

Automated 10-day sweep (2026-07-08 → 2026-07-18) of closed PRs in espg/mortie for deferred/follow-up work.

Only 5 PRs closed in this window — #102, #105, #106, #109, #110 — all closed on 2026-07-10. No PR has closed since. All 5 were already swept by last week's routine and filed in #115 ("Follow up items for the week of July 11, 2026"), which is still open, still labeled waiting on espg, and has had no activity since a detailed options comment was posted on 2026-07-13.

So there's no new PR-closure activity to report this week. This issue re-verifies #115's two items against current main and adds one item #115 missed.

Re-verified from #115 (both still open, both still accurate against current main)

1. Vertex-point-touch closed-set gap in edge_hits_cell_edge — from PR Uniform symbolic crossing predicate: fix base-cell-boundary mis-fill #106, body Question 4. Confirmed the documenting comment at src_rust/src/coverage.rs:524-531 is unchanged since PR Gate the seed cull to sub-hemisphere caps #110 merged (checked against origin/main).
2. MortonIndexArray.__getitem__ breaks the pandas ExtensionArray NA contract — from PR Decimal-string display & casting layer for morton_index #105 review comment at morton_index.py:351. Confirmed mortie/morton_index.py:348-352 still returns MortonIndexScalar(result) unconditionally, no sentinel special-case.

Both are already tracked with full options analysis on #115 — not restating the detail here, just confirming they're still live.

New item (not previously tracked)

3. No regression test pins the sub-hemisphere donor-cull path against unculled classification — PR #109 review comment (3) asked for a fuzz-style test asserting base_fills agrees with unculled classification over a family that exercises the normal (≤90°) cull-skip/donor-chain mechanism, with a ready-made 360-config fuzz family (grid-aligned boxes with one edge on the equator) offered. PR Gate the seed cull to sub-hemisphere caps #110 added test_base_fills_oversize_cap_classifies_all_seeds, but that pins the oversize (>90°) cap path where the cull disengages — a different branch from what was asked. The suggested fuzz test does not exist in src_rust/src/coverage/tests.rs.

Priority table

#	Item	Source	Severity
1	Vertex-point-touch closed-set gap (neighbor-cell case)	#106	Medium
2	MortonIndexArray.__getitem__ NA-contract violation	#105	Medium
3	Missing donor-cull fuzz test	#109	Low

Recommendation

Items (1) and (2) are already being tracked on #115 with more detail than restating here would add — suggest folding item (3) into #115 as an addendum rather than tracking it in parallel, to avoid duplicate issue-tracking for the same PR window. This issue is filed per the routine's weekly-issue convention; happy to close it in favor of #115 if that's preferred.

---

Automated follow-up sweep — see CLAUDE.md §5 for the discuss label workflow this issue expects.