feat(registry-webhook-sync): Introducing Terraform Registry webhook sync workflow 🚀

[clouddrove-ci]

Problem

New tags for terraform-aws-* modules sometimes don't appear on the public Terraform Registry, forcing a manual Resync Module click in the registry UI. This automates the fix and makes it self-healing across orgs.

Root cause (investigated, with live evidence)

The registry installs a GitHub webhook per published repo. On every clouddrove module repo that hook is subscribed to push only:

• git push origin <tag> (our release-tag.yml) fires a push event → registry syncs ✅
• A tag cut via the GitHub Release UI/API fires create + release, often not push → the push-only hook never hears it → registry lags → manual resync ⚠️

There is no public registry API to force a resync (UI-only; the v2 can-resync action needs a fragile registry session token). So the durable fix is correcting the GitHub webhook, not calling the registry.

Evidence gathered (read-only, against clouddrove)

1. Org-wide webhook audit (dry-run, 177 terraform-* repos):

Result	Count
Hook = ["push"] only → would fix to ["push","create","delete"]	120
No registry hook at all (unpublished / hook deleted)	57
Already correct	0

2. git-tag vs registry-version spot check (v-prefix normalized):

Repo	git latest	registry latest	state
terraform-aws-eks	1.4.8	1.4.8	SYNCED
terraform-aws-vpc	2.0.5	2.0.5	SYNCED
terraform-aws-ec2	2.1.1	2.1.1	SYNCED
terraform-aws-iam	4.11.0	4.11.0	SYNCED
terraform-aws-kms	1.3.4	1.3.4	SYNCED
terraform-aws-s3	2.0.2	2.0.1	LAG
terraform-aws-lambda	1.3.3	1.3.2	LAG

So most modules sync fine (push-driven); the misses are intermittent and align with manually-cut releases (e.g. s3 v2.0.2 was a hand-cut GitHub Release). GitHub had purged old webhook delivery history, so the exact missed delivery couldn't be replayed — diagnosis rests on the audit + lag correlation + HashiCorp's documented requirement that the registry webhook fire on tag-creation events.

3. Reconcile logic validated against live repos in dry-run — drift detection correctly reports have=["push"] missing=["create","delete"] per repo and makes no changes in dry_run mode.

What this PR adds

• .github/workflows/registry-webhook-sync.yml — reusable (workflow_call) + workflow_dispatch. Loops orgs → terraform-* repos → finds the registry hook → adds missing create/delete events. Has dry_run, writes a job-summary report, leaves repos with no registry hook untouched (reported only).
• docs/registry-webhook-sync.md — problem writeup, inputs/secrets, scheduled cron caller example.
• WORKFLOW_CATALOG.md — catalog entry.

Inputs: orgs (csv, required), repo_filter (default ^terraform-), required_events (default push,create,delete), dry_run. Secret: GH_ADMIN_TOKEN (PAT with admin:repo_hook + repo read).

Honest scope / caveats

• Fixes future tags. Versions already missed need one manual resync each to backfill.
• The webhook fix is probabilistic (adds a second trigger). For a guaranteed catch, a follow-up safety-net workflow can verify the version landed post-release and Slack-alert if not — not in this PR.
• 57 repos have no registry hook — likely unpublished modules; only the registry creates that hook (on publish). Separate triage.

Validation done

• YAML parses; has name: + jobs: (passes ci.yml structure checks).
• Core reconcile loop executed live in dry_run against clouddrove — results above.

Follow-ups (not in this PR)

• Run live on clouddrove (apply the 120) after review.
• Add scheduled caller (cron) with all 5 org names in an ops repo.
• Optional safety-net verify+notify workflow.
• Triage the 57 no-hook repos.