feat(experimentation): refresh experiment exposures via task processor

[gagantrivedi]

• I have read the Contributing Guide.
• I have added information to docs/ if required so people know about the feature. (deferred — flag-gated and not user-visible yet; docs land with the UI. Events catalogue regenerated here.)
• I have filled in the "Changes" section below.
• I have filled in the "How did you test this code" section below.

Changes

Contributes to the experimentation results layer (v0.1: experiment exposures panel). Second backend slice, stacked on #7740 — the API endpoint and UI follow.

• compute_experiment_exposures(experiment_id) — computes the full-window exposures summary and stores it via record_refresh. Window is started_at → ended_at or now, so a completed experiment's final refresh is frozen at its end. Warehouse errors are absorbed into record_failure (the last good payload survives; nothing raises into the task processor). Never-started experiments are skipped.
• refresh_running_experiment_exposures — recurring sweep every EXPERIMENTATION_EXPOSURES_REFRESH_MINUTES (new setting, default 60) enqueueing one compute per RUNNING experiment; no-op when EXPERIMENTATION_CLICKHOUSE_URL is unset. No staleness filter: the sweep cadence is the refresh interval, and an as_of-older-than-interval check would skip every other sweep (task latency keeps as_of slightly younger than the interval).
• Completing an experiment enqueues the final refresh from transition_experiment_status, making the stored row the durable record once events pass the 90-day ClickHouse TTL.
• structlog events — experimentation.exposures.computed / exposures.compute_failed; events catalogue regenerated.

How did you test this code?

• Unit tests: compute task (stored payload/as_of, completed-experiment frozen window, failure preserves last good payload, never-started skip, log events asserted), sweep (enqueues running experiments only; skips without a warehouse DSN), complete action enqueues the final refresh / start does not.
• pytest tests/unit/experimentation/ — 257 passed; ruff + mypy strict clean.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	Jun 11, 2026 8:25am

2 Skipped Deployments

Project	Deployment	Actions	Updated (UTC)
flagsmith-frontend-preview	Ignored	Preview	Jun 11, 2026 8:25am
flagsmith-frontend-staging	Ignored	Preview	Jun 11, 2026 8:25am

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.55%. Comparing base (ef36c14) to head (7aac101).

Additional details and impacted files

@@                         Coverage Diff                         @@
##           feat/experiment-exposure-snapshots    #7751   +/-   ##
===================================================================
  Coverage                               98.55%   98.55%           
===================================================================
  Files                                    1454     1453    -1     
  Lines                                   55987    56086   +99     
===================================================================
+ Hits                                    55177    55276   +99     
  Misses                                    810      810           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[gagantrivedi]

Superseded: the task-processor approach is cut in favour of an explicit refresh endpoint, landing with the GET endpoint in the next PR.