Fetch Throughput & Small Server Defaults - Source Monitor

UAT Tests

Scenario: Verify error handling safety net prevents sources from getting permanently stuck in "fetching" status.

Expected: DB errors propagate (not swallowed), broadcast errors are rescued, ensure block exists as safety net.

Result: PASS

update_source_state! (fetch_runner.rb:90-100): source.update!(attrs) is bare (no rescue), broadcast wrapped in begin/rescue
run method (lines 72-78): ensure block reloads source, resets "fetching" to "failed", defensive inner rescue
FollowUpHandler#call (follow_up_handler.rb:19-25): per-item begin/rescue wraps each enqueue call with logging

Scenario: Verify fixed-interval path uses jitter instead of exact intervals.

Expected: Fixed-interval path uses adjusted_interval_with_jitter(fixed_seconds) for +-10% jitter.

Result: PASS

adaptive_interval.rb lines 31-33: fixed_seconds = fixed_minutes * 60.0, then adjusted_interval_with_jitter(fixed_seconds) — no longer exact minutes

Scenario: Verify scheduler defaults are small-server friendly and configurable.

Expected: batch_size=25, stale_timeout=5, both configurable via DSL.

Result: PASS

fetching_settings.rb: @scheduler_batch_size = 25, @stale_timeout_minutes = 5 in reset!
scheduler.rb reads from config with legacy constant fallbacks
stalled_fetch_reconciler.rb reads from config with 10.minutes fallback

Scenario: Verify queue separation: 7 non-fetch jobs on maintenance, only 2 jobs on fetch.

Expected: 3 queues (fetch, scrape, maintenance), proper job assignments, example config updated.

Result: PASS

configuration.rb: maintenance_queue_name defaults to "source_monitor_maintenance", concurrency 1
queue_name_for(:maintenance) and concurrency_for(:maintenance) both work
7 jobs on :maintenance: source_health_check, import_session_health_check, import_opml, log_cleanup, item_cleanup, favicon_fetch, download_content_images
2 jobs on :fetch: fetch_feed, schedule_fetches
solid_queue.yml has all 3 queues with env var overrides and explanatory comments