fix(docs): correct the PR template's closing-keyword note (backend#2188)

[saqlainsyed007]

Closes tracebloc/backend#2188

Problem

The PR template's line-5 comment carried:

PRs land on develop, not the default branch, so closing keywords do not fire on merge — confirm the issue actually closed.

Both halves are false: developis this repo's default branch, so a same-repo closing keyword does fire on merge (measured 12/12 on recent base:develop merges in backend). A wrong doc is worse than an unhelpful one now that agents read docs first — this exact sentence manufactured a High-severity Bugbot finding on e2e-test-agent#165.

Fix

Replace only the false sentence with the accurate one (as landed in e2e-test-agent#165). The load-bearing cross-repo half — a bare repo#N closes nothing, only an owner-qualified tracebloc/repo#N — is preserved, along with this repo's own example.

One of 8 repos carrying the clause (backend#2188): backend, client, averaging-service, frontend-app, tracebloc-py-package, data-ingestors, design-system, cli.

---

Note

Low Risk
Documentation-only change to a PR template comment; no runtime, security, or deployment impact.

Overview
Updates the Related section comment in .github/pull_request_template.md so it no longer claims that PRs merge to a non-default branch and that same-repo closing keywords do not work on merge.

The revised text states that develop is this repo's default branch, so same-repo keywords like Closes #123do fire on merge (with a note that this was verified 12/12). The existing guidance on owner-qualified cross-repo references (tracebloc/client-runtime#456 vs a bare client-runtime#456) is unchanged.

^{Reviewed by Cursor Bugbot for commit 14d2633. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

Blocking on one clause, shared by all nine copies. Full evidence on the canonical, tracebloc/backend#2717.

Short version: the half you fixed is right — develop is the default branch and a same-repo Closes #123 does fire. But the replacement adds a new falsehood:

"A cross-repo one never does, whatever the branch: close those by hand."

An owner-qualified cross-repo keyword fires too. Measured on two merges from today:

e2e-test-agent#291 "Closes tracebloc/backend#2693" merged 14:24:02Z -> #2693 CLOSED 14:24:03Z
tracebloc-py-package#463 -> tracebloc/backend#2646 merged 11:13:49Z -> #2646 CLOSED 11:13:51Z

The real rule is owner-qualification, not same-repo vs cross-repo — which is exactly what the clause you correctly kept already says ("a bare <repo>#456 closes nothing"). The new sentence contradicts it and removes the only reason to owner-qualify.

I'd drop the sentence rather than reword it. CI is green here, so this clause is the only thing holding it.

[saqlainsyed007]

Dropped the false cross-repo one never does clause here too — full context on the canonical backend#2717. Re-requesting review.

[LukasWodka]

Approving — the false clause is gone and what's left is accurate.

I re-extracted the comment line from all nine heads and hashed it: three hashes, differing only in the per-repo example (client-runtime / frontend-app / backend), and "A cross-repo one never does, whatever the branch" is absent from every one. That was the whole blocker.

What stands now is true on each claim: Closes #123 fires because develop is the default branch; a bare<repo>#456 closes nothing because it isn't a reference; owner-qualifying is what makes a cross-repo link real. Dropping the sentence rather than rewriting it was the better call — the parenthetical already carried the rule.

One non-blocking note, offered once and then I'll leave it. The line now states the same-repo case and says nothing about whether an owner-qualified cross-repo keyword fires. It does — I measured it earlier today, e2e-test-agent#291 merged 14:24:02Z and backend#2693 closed 14:24:03Z. Since the entire ticket exists because people believed keywords don't fire, silence on that half is the one place a reader can still guess wrong. "Both fire on merge" would close it. Your call entirely — omission isn't the falsehood I was blocking, and this is already a clear improvement on what was there.

Good turnaround on nine repos.