spike(aidd-vcs): establish trustworthy related-delivery signals

No files or tests are named. Start by inventorying the configured VCS support and creating fixtures for open related PRs, merged work, unrelated branches, and branches without issue numbers. Done means a signal matrix, evidence-preserving result vocabulary, stable identities or URLs for confirmed signals, and a decision on whether a narrowly scoped VCS pre-flight contract is justified.

Question

Which delivery signals can reliably tell a backlog workflow that remote work is related to an issue, without treating an absent naming match as evidence that no work exists?

Current constraint

The project branch convention is type/ticket-short-description; it does not require an issue number. The previous proposal treated an absent feat/N-* branch as clear, which can be a false negative. Remote branch and pull-request observation belongs to VCS; backlog capabilities consume the resulting evidence.

Scope

• Inventory the related-delivery signals a configured VCS support can expose, including explicit issue-to-PR links, merged delivery, branch references, and ambiguous textual matches.
• Test the evidence quality of each signal against fixtures with an open related PR, merged related work, an unrelated branch, and a branch without an issue number.
• Define an evidence-preserving result vocabulary. It must distinguish confirmed active work, confirmed landed work, no known related signal, and unknown or ambiguous state.
• Decide the boundary between VCS observation and PM/backlog consumption.

Evidence of completion

• A written signal matrix records each tested source, its confidence, and the result it may justify.
• No result treats an absent branch-name match as proof that no related work exists.
• The output names the stable identity or URL for every confirmed signal.
• The decision states whether a VCS pre-flight feature is justified and, only then, its narrowest contract.

Out of scope

• Implementing a watcher, issue refresh, auto-rebase, or branch/worktree creation.
• Semantic code-drift analysis.
• Fuzzy matching presented as confirmed related work.