chore(warehouse-sources): route v2 pipeline post-load through shared orchestrator

[estefaniarabadan]

Problem

The v2 pipeline's _post_run_operations carried an inline copy of the shared post-load sequence in pipelines/common/load.py — compaction, S3 publishing, sync bookkeeping, table validation, CDC companion seeding, view syncing. Every fix had to be written twice: #74630 is a recent example, patching the same transient-S3 handling in both copies in one PR. And the shared orchestrator itself still mixed the core publish/register flow with CDC-specific branching and product side effects.

This is part 2 of 2 of the post-load unification (following #74671). With this change, pipelines/common/ has a single post-load implementation used by both pipeline versions and CDC companion seeding, with the CDC-specific logic and product side effects behind their own seams.

Changes

Commit 1 — route v2 through the shared orchestrator:

PipelineNonDLT._post_run_operations is now a thin wrapper calling run_post_load_operations, forwarding the resource and last incremental value (so desc-sort cursor finalization keeps working) and returning the orchestrator's queryable folder (so ducklake's prepared_queryable_folder keeps flowing). The ~95-line duplicated block and six now-unused imports are deleted. The wrapper stays a method because the e2e tests patch.object it as a seam.

Commit 2 — finish the orchestrator's step shape:

• _register_tables_and_seed is split into _register_table (plain table validation/registration) and _run_cdc_post_load (companion registration + snapshot seeding), the latter invoked only for CDC schemas and companion writes — so the non-CDC path no longer routes through CDC branching.
• The revenue-ready notification, revenue/engineering analytics view syncs, and repartition detection move into POST_LOAD_STEPS, a tuple of callables sharing one PostLoadStep protocol signature. This is the seam where product hooks can later register post-load side effects (following the external_product_hooks.py pattern) instead of editing this module.

Note

Behavior deltas, each aligning v2 with what v3 already does or reordering independent operations:

• v2's CDC-snapshot runs get threshold-based delta maintenance instead of unconditional compaction
• v2 gains post-load repartition detection (feature-flag gated, never raises) and post-load duration metrics
• v2's existing_queryable_folder is read live from schema.table instead of the pipeline-start snapshot (fresher, and correct when partial loading created the table mid-run)
• the revenue-ready notification now runs after table registration instead of between last_synced_at and initial_sync_complete (the operations are independent)
• the cdc_seed_check info log no longer fires for non-CDC schemas

How did you test this code?

• pipelines/pipeline_v2/ + pipelines/common/: all passing after each commit (43 and 41+15 respectively)
• Repo-wide uv run mypy --cache-fine-grained .: no issues in 17,083 files
• ruff check / ruff format: clean

Test changes: deleted TestPostRunOperationsCompactionErrorHandling — it exercised the deleted duplicated block, and the same regressions (transient S3 blip vs genuine failure, for both the compact and maintenance paths) are covered by pipelines/common/test/test_load.py. Added one test asserting the wrapper returns the orchestrator's queryable folder and forwards resource/last_incremental_field_value — that regression is otherwise silent: ducklake registration reads the result with .get(...), desc-sort finalization just skips, and every e2e test mocks _post_run_operations, so no existing test would notice. The commit-2 restructure adds no new branches, so no new tests.

Automatic notifications

• Publish to changelog?
• Alert Sales and Marketing teams?

Docs update

No user-facing or documented-workflow changes.

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Authored with Claude Code as phase 2 (part 2 of 2) of the warehouse pipelines separation plan, continuing #74352 and #74671. Skills invoked: /writing-tests (gating the test deletion and the one new test). Verification: repo-wide mypy, targeted hogli test runs, ruff. Decisions: kept _post_run_operations as a method rather than inlining into run() because the e2e suite patches it via patch.object; POST_LOAD_STEPS is a plain module-level tuple rather than a registry — the registration mechanism is deliberately out of scope per the plan, this commit only shapes the entries to a shared signature; the behavior deltas listed above were accepted deliberately rather than parameterized away, since each matches the v3 path the orchestrator already serves.

[trunk-io]

😎 Merged successfully - details.

[github-actions]

Hey @estefaniarabadan! 👋

It looks like your git author email on this PR isn't your @posthog.com address (estefania.rabadan@gmail.com). Since you're on the PostHog team, it's worth pointing your local git author email at your @posthog.com address. Why it matters:

• Consistent work identity in git history — internal tooling that attributes commits to team members keys off your @posthog.com address.
• Keeps team contributions easy to tell apart from external community ones when scanning history.

You can fix it for this repo with:

git config user.email "you@posthog.com"

Or set it globally with git config --global user.email "you@posthog.com". No need to redo this PR — just a nudge for next time. 🙂

[greptile-apps]

_{Reviews (1): Last reviewed commit: "chore(warehouse-sources): route v2 pipel..." | Re-trigger Greptile}

[danielcarletti]

YES

[estefaniarabadan]

/trunk merge

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-07-30 11:59 UTC	Run
prod-us	✅ Deployed	2026-07-30 12:17 UTC	Run
prod-eu	✅ Deployed	2026-07-30 12:22 UTC	Run