Say what the code does now, not what it did before

commit-branch.sh

@@ -47,10 +47,9 @@ cd "$WORKDIR"
 
 # createCommitOnBranch commits onto a branch that already exists; it cannot
 # create one, because a branch needs a commit and this is how commits are made.
-# The old git-push path bootstrapped a missing branch with `git init`, and that
-# is the one thing lost here. Both state branches have existed since the
-# archive did, so this fires only if one is deleted -- in which case the fix is
-# to recreate it deliberately, not to have a workflow guess at its contents.
+# Bootstrapping a missing branch is deliberately not done here: this fires only
+# if a state branch is deleted, and the fix then is to recreate it deliberately
+# rather than have a workflow guess at its contents.
 if ! git rev-parse --git-dir >/dev/null 2>&1; then
     log "FATAL: $WORKDIR is not a checkout of $BRANCH."
     log "       The branch has to exist before anything can be committed to it."

tests/run.sh

@@ -151,10 +151,10 @@ echo "an HTTP failure reports what GitHub said, not just curl's exit code"
     git -c user.name=t -c user.email=t@example.invalid commit -qm base
     printf 'two\n' > keep.txt
 
-    # curl as --fail-with-body behaves on an HTTP error: the body is written
-    # to the output AND the exit status is non-zero. That combination is what
-    # used to lose the message -- set -e took the exit before anything printed
-    # the body, and the trap then deleted the file.
+    # curl as --fail-with-body behaves on an HTTP error: the body is written to
+    # the output AND the exit status is non-zero. That combination loses the
+    # message unless it is handled -- set -e takes the exit before anything
+    # prints the body, and the trap then deletes the file.
     mkdir -p "$work/bin"
     cat > "$work/bin/curl" <<'FAKE'
 #!/bin/sh