docs: TODO deferred merge-stats split and squash-preview coverage

src/commands/step/squash.rs

@@ -370,6 +370,10 @@ fn preview_squash(target: Option<&str>, dry_run: bool, yes: bool) -> anyhow::Res
 
     let integration_target = repo.require_target_ref(target)?;
     // #3519: preview against the same upstream-aware span the real squash uses.
+    // TODO(#3519 follow-up): unlike `handle_squash`, this path (--dry-run /
+    // --show-prompt) has no test asserting it measures against the stale
+    // target's upstream — a snapshot test pinning the preview output in that
+    // topology would close the one unasserted consumer of `span_upstream`.
     let span_target = repo
         .span_upstream(&integration_target)?
         .unwrap_or(integration_target);

src/commands/worktree/push.rs

@@ -124,6 +124,13 @@ impl MergeContext {
         let stash_guard =
             repo.prepare_target_worktree(target_worktree_path.as_ref(), &target_branch)?;
 
+        // TODO(#3519 follow-up): when `target_branch` was behind its upstream
+        // (see `Repository::span_upstream`), this count mixes the carried
+        // fast-forwarded upstream commits in with the branch's own squash
+        // commit, so the success line ("Merged to main (N commits, ...)")
+        // overstates what the branch itself contributed. Splitting them needs
+        // a carried-count threaded through `MergeContext`; deferred as
+        // cosmetic.
         let commit_count = repo.count_commits(&target_branch, "HEAD")?;
 
         let stats_summary = if commit_count > 0 {