diff --git a/inc/Steps/EventImport/Handlers/EventFlyer/EventFlyer.php b/inc/Steps/EventImport/Handlers/EventFlyer/EventFlyer.php
index 904dfdc..3446946 100644
--- a/inc/Steps/EventImport/Handlers/EventFlyer/EventFlyer.php
+++ b/inc/Steps/EventImport/Handlers/EventFlyer/EventFlyer.php
@@ -12,6 +12,7 @@
 namespace DataMachineEvents\Steps\EventImport\Handlers\EventFlyer;
 
 use DataMachine\Core\ExecutionContext;
+use DataMachine\Core\Steps\Fetch\FreshCandidateCollector;
 use DataMachineEvents\Steps\EventImport\Handlers\EventImportHandler;
 use DataMachineEvents\Steps\EventImport\Handlers\VenueFieldsTrait;
 use DataMachineEvents\Utilities\EventIdentifierGenerator;
@@ -140,6 +141,12 @@ private function getNextUnprocessedImage( ExecutionContext $context ): ?array {
 
 		$image_extensions = array( 'jpg', 'jpeg', 'png', 'gif', 'webp' );
 
+		// Selection-time prefilter via Data Machine core primitive.
+		// FreshCandidateCollector skips already-processed and currently-claimed
+		// images consistently with FetchHandler's authoritative dedup. Authoritative
+		// dedup/claim/cap still runs in FetchHandler::get_fetch_data() after this.
+		$collector = new FreshCandidateCollector( $context, 1 );
+
 		foreach ( $repo_files as $file ) {
 			$extension = strtolower( pathinfo( $file['filename'], PATHINFO_EXTENSION ) );
 
@@ -148,23 +155,25 @@ private function getNextUnprocessedImage( ExecutionContext $context ): ?array {
 			}
 
 			$file_identifier = $file['path'];
-
-			// Pre-filter: skip already-processed images to find the next one.
-			// This is a selection mechanism, not dedup — dedup happens in FetchHandler::dedup().
-			if ( $context->isItemProcessed( $file_identifier ) ) {
-				continue;
-			}
-
-			return array(
+			$image           = array(
 				'original_name'   => $file['filename'],
 				'persistent_path' => $file['path'],
 				'size'            => $file['size'],
 				'mime_type'       => $this->getMimeType( $file['path'] ),
 				'uploaded_at'     => gmdate( 'Y-m-d H:i:s', $file['modified'] ),
 			);
+
+			$collector->offer( $file_identifier, $image );
+
+			if ( $collector->isFull() ) {
+				break;
+			}
 		}
 
-		return null;
+		$collector->markExhausted();
+
+		$accepted = $collector->getAccepted();
+		return $accepted[0] ?? null;
 	}
 
 	private function getMimeType( string $file_path ): string {
diff --git a/inc/Steps/EventImport/Handlers/EventImportHandler.php b/inc/Steps/EventImport/Handlers/EventImportHandler.php
index 8a59e25..1da29d8 100644
--- a/inc/Steps/EventImport/Handlers/EventImportHandler.php
+++ b/inc/Steps/EventImport/Handlers/EventImportHandler.php
@@ -110,39 +110,6 @@ public function isPastEvent( string $start_date ): bool {
 		return strtotime( $start_date ) < strtotime( 'today' );
 	}
 
-	/**
-	 * Check if item has been processed (uses ExecutionContext).
-	 *
-	 * @deprecated Use metadata['item_identifier'] instead. Dedup is now centralized in FetchHandler::dedup().
-	 *
-	 * @param ExecutionContext $context Execution context
-	 * @param string           $item_identifier Item identifier
-	 * @return bool True if already processed
-	 */
-	public function checkItemProcessed( ExecutionContext $context, string $item_identifier ): bool {
-		return $context->isItemProcessed( $item_identifier );
-	}
-
-	/**
-	 * Mark item as processed (uses ExecutionContext).
-	 *
-	 * Also stores item context in engine data for the skip_item tool.
-	 *
-	 * @deprecated Use metadata['item_identifier'] instead. Dedup is now centralized in FetchHandler::dedup().
-	 *
-	 * @param ExecutionContext $context Execution context
-	 * @param string           $item_identifier Item identifier
-	 */
-	public function markItemAsProcessed( ExecutionContext $context, string $item_identifier ): void {
-		$context->markItemProcessed( $item_identifier );
-
-		// Store item context for skip_item tool
-		$job_id = $context->getJobId();
-		if ( $job_id ) {
-			EventEngineData::storeItemContext( (int) $job_id, $item_identifier, $this->handler_type );
-		}
-	}
-
 	/**
 	 * Called by FetchHandler::dedup() after marking an item as processed.
 	 *
diff --git a/inc/Steps/EventImport/Handlers/WebScraper/EventSectionFinder.php b/inc/Steps/EventImport/Handlers/WebScraper/EventSectionFinder.php
index be69eca..a82997f 100644
--- a/inc/Steps/EventImport/Handlers/WebScraper/EventSectionFinder.php
+++ b/inc/Steps/EventImport/Handlers/WebScraper/EventSectionFinder.php
@@ -11,6 +11,7 @@
 namespace DataMachineEvents\Steps\EventImport\Handlers\WebScraper;
 
 use DataMachine\Core\ExecutionContext;
+use DataMachine\Core\Steps\Fetch\FreshCandidateCollector;
 
 if ( ! defined( 'ABSPATH' ) ) {
 	exit;
@@ -19,9 +20,9 @@
 final class EventSectionFinder {
 
 	/**
-	 * @var callable(string): bool
+	 * @var FreshCandidateCollector
 	 */
-	private $is_item_processed;
+	private FreshCandidateCollector $fresh_candidates;
 
 	/**
 	 * @var callable(string): string
@@ -34,12 +35,12 @@ final class EventSectionFinder {
 	private $is_past_event;
 
 	/**
-	 * @param callable(string): bool $is_item_processed Check if item identifier has been processed
+	 * @param FreshCandidateCollector $fresh_candidates Selection-time fresh candidate collector.
 	 * @param callable(string): string $clean_html_for_ai Clean HTML for AI processing
 	 * @param callable(string): bool $is_past_event Check if Y-m-d date is in the past
 	 */
-	public function __construct( callable $is_item_processed, callable $clean_html_for_ai, callable $is_past_event ) {
-		$this->is_item_processed = $is_item_processed;
+	public function __construct( FreshCandidateCollector $fresh_candidates, callable $clean_html_for_ai, callable $is_past_event ) {
+		$this->fresh_candidates  = $fresh_candidates;
 		$this->clean_html_for_ai = $clean_html_for_ai;
 		$this->is_past_event     = $is_past_event;
 	}
@@ -96,7 +97,7 @@ public function find_first_eligible_section( string $html_content, string $url,
 					// Build stable identifier from decoded data
 					$event_identifier = md5( $url . $summary . ( $event_data['start'] ?? '' ) );
 
-					if ( call_user_func( $this->is_item_processed, $event_identifier ) ) {
+					if ( ! $this->fresh_candidates->offer( $event_identifier ) ) {
 						continue;
 					}
 
@@ -128,12 +129,12 @@ public function find_first_eligible_section( string $html_content, string $url,
 				$content_hash     = md5( $raw_html );
 				$event_identifier = md5( $url . $content_hash );
 
-				if ( call_user_func( $this->is_item_processed, $event_identifier ) ) {
+				$cleaned_html = call_user_func( $this->clean_html_for_ai, $raw_html );
+				if ( empty( $cleaned_html ) || strlen( $cleaned_html ) < 30 ) {
 					continue;
 				}
 
-				$cleaned_html = call_user_func( $this->clean_html_for_ai, $raw_html );
-				if ( empty( $cleaned_html ) || strlen( $cleaned_html ) < 30 ) {
+				if ( ! $this->fresh_candidates->offer( $event_identifier ) ) {
 					continue;
 				}
 
diff --git a/inc/Steps/EventImport/Handlers/WebScraper/UniversalWebScraper.php b/inc/Steps/EventImport/Handlers/WebScraper/UniversalWebScraper.php
index 30b3be3..90b4553 100644
--- a/inc/Steps/EventImport/Handlers/WebScraper/UniversalWebScraper.php
+++ b/inc/Steps/EventImport/Handlers/WebScraper/UniversalWebScraper.php
@@ -48,6 +48,7 @@
 namespace DataMachineEvents\Steps\EventImport\Handlers\WebScraper;
 
 use DataMachine\Core\ExecutionContext;
+use DataMachine\Core\Steps\Fetch\FreshCandidateCollector;
 use DataMachineEvents\Steps\EventImport\Handlers\EventImportHandler;
 use DataMachineEvents\Steps\EventImport\Handlers\WebScraper\Extractors\ExtractorInterface;
 use DataMachineEvents\Steps\EventImport\Handlers\WebScraper\Extractors\WixEventsExtractor;
@@ -469,12 +470,18 @@ private function tryHtmlSectionExtraction(
 		ExecutionContext $context,
 		int $current_page
 	): ?array {
-		$skipped_identifiers = array();
+		// Selection-time prefilter via Data Machine core primitive.
+		// The collector skips processed/claimed/duplicate sections while this
+		// method keeps event-specific filters (HTML validity, title keywords,
+		// include/exclude keywords). FetchHandler remains authoritative after
+		// executeFetch() returns.
+		$section_collector = new FreshCandidateCollector( $context );
 
 		while ( true ) {
-			$event_section = $this->extract_event_sections( $html_content, $current_url, $context, $skipped_identifiers );
+			$event_section = $this->extract_event_sections( $html_content, $current_url, $context, $section_collector );
 
 			if ( empty( $event_section ) ) {
+				$section_collector->markExhausted();
 				break;
 			}
 
@@ -490,13 +497,11 @@ private function tryHtmlSectionExtraction(
 			$raw_html_data = $this->extract_raw_html_section( $event_section['raw_html'], $current_url, $context, $config );
 
 			if ( ! $raw_html_data ) {
-				$skipped_identifiers[ $event_section['identifier'] ] = true;
 				continue;
 			}
 
 			$section_title = $this->extract_section_title( $raw_html_data );
 			if ( '' !== $section_title && $this->shouldSkipEventTitle( $section_title ) ) {
-				$skipped_identifiers[ $event_section['identifier'] ] = true;
 				continue;
 			}
 
@@ -511,7 +516,6 @@ private function tryHtmlSectionExtraction(
 						'source_url'         => $current_url,
 					)
 				);
-				$skipped_identifiers[ $event_section['identifier'] ] = true;
 				continue;
 			}
 
@@ -524,7 +528,6 @@ private function tryHtmlSectionExtraction(
 						'source_url'         => $current_url,
 					)
 				);
-				$skipped_identifiers[ $event_section['identifier'] ] = true;
 				continue;
 			}
 
@@ -740,11 +743,11 @@ private function fetch_html( string $url, ExecutionContext $context ): string {
 	}
 
 	/**
-	 * Extract first non-processed event HTML section from content.
+	 * Extract first fresh event HTML section from content.
 	 */
-	private function extract_event_sections( string $html_content, string $url, ExecutionContext $context, array $skipped_identifiers = array() ): ?array {
+	private function extract_event_sections( string $html_content, string $url, ExecutionContext $context, FreshCandidateCollector $section_collector ): ?array {
 		$finder = new EventSectionFinder(
-			fn ( string $identifier ): bool => isset( $skipped_identifiers[ $identifier ] ) || $context->isItemProcessed( $identifier ),
+			$section_collector,
 			fn ( string $html ): string => $this->clean_html_for_ai( $html ),
 			fn ( string $ymd ): bool => $this->isPastEvent( $ymd )
 		);
diff --git a/inc/Steps/EventImport/Handlers/WebScraper/VisionExtractionProcessor.php b/inc/Steps/EventImport/Handlers/WebScraper/VisionExtractionProcessor.php
index f28f597..6ac644b 100644
--- a/inc/Steps/EventImport/Handlers/WebScraper/VisionExtractionProcessor.php
+++ b/inc/Steps/EventImport/Handlers/WebScraper/VisionExtractionProcessor.php
@@ -13,6 +13,7 @@
 namespace DataMachineEvents\Steps\EventImport\Handlers\WebScraper;
 
 use DataMachine\Core\ExecutionContext;
+use DataMachine\Core\Steps\Fetch\FreshCandidateCollector;
 use DataMachineEvents\Steps\EventImport\Handlers\EventImportHandler;
 use DataMachineEvents\Steps\EventImport\Handlers\WebScraper\Extractors\VisionExtractor;
 use DataMachineEvents\Steps\EventImport\EventEngineData;
@@ -76,41 +77,57 @@ public function process(
 			)
 		);
 
+		// Selection-time prefilter via Data Machine core primitive.
+		// FreshCandidateCollector skips already-processed and currently-claimed
+		// images consistently with FetchHandler's authoritative dedup. Authoritative
+		// dedup/claim/cap still runs in FetchHandler::get_fetch_data() after this.
+		$collector = new FreshCandidateCollector( $context, 1 );
 		foreach ( $candidates as $candidate ) {
-			$image_url = $candidate['url'];
-
-			// Generate content-based identifier for cross-run tracking.
+			$image_url        = $candidate['url'];
+			// Content-based identifier for cross-run tracking.
 			$image_identifier = md5( $url . $image_url );
 
-			// Pre-filter: skip already-processed images to find the next candidate.
-			// This is a selection mechanism, not dedup — dedup happens in FetchHandler::dedup().
-			if ( $context->isItemProcessed( $image_identifier ) ) {
-				$context->log(
-					'debug',
-					'VisionExtractor: Skipping processed image',
-					array( 'image_url' => $image_url )
-				);
-				continue;
+			$collector->offer(
+				$image_identifier,
+				array(
+					'image_url'        => $image_url,
+					'image_identifier' => $image_identifier,
+					'score'            => $candidate['score'] ?? 0,
+				)
+			);
+
+			if ( $collector->isFull() ) {
+				break;
 			}
+		}
+		$collector->markExhausted();
+
+		$accepted = $collector->getAccepted();
+		if ( ! empty( $accepted ) ) {
+			$selected         = $accepted[0];
+			$image_url        = $selected['image_url'];
+			$image_identifier = $selected['image_identifier'];
 
 			$context->log(
 				'debug',
 				'VisionExtractor: Processing image candidate',
 				array(
 					'image_url' => $image_url,
-					'score'     => $candidate['score'],
+					'score'     => $selected['score'],
 				)
 			);
 
 			// Download to persistent storage.
 			$file_path = $this->downloadImageToPersistentStorage( $image_url, $context );
 
-			// Mark as processed AFTER download attempt (success or fail).
-			// Direct call — this is a pre-filter, not dedup. Centralized dedup
-			// in FetchHandler::dedup() runs after executeFetch() returns.
-			$context->markItemProcessed( $image_identifier );
-
 			if ( ! $file_path ) {
+				// Mark this candidate as processed even though download failed,
+				// so the next run advances to the next candidate instead of
+				// retrying a permanently broken URL. This is a retry-prevention
+				// concern separate from candidate prefiltering - the collector
+				// only reads processed state, it does not mark.
+				$context->markItemProcessed( $image_identifier );
+
 				$context->log(
 					'warning',
 					'VisionExtractor: Failed to download image, will try next candidate on next run',
diff --git a/tests/Unit/EventSectionFinderTest.php b/tests/Unit/EventSectionFinderTest.php
new file mode 100644
index 0000000..065fcb4
--- /dev/null
+++ b/tests/Unit/EventSectionFinderTest.php
@@ -0,0 +1,59 @@
+<?php
+/**
+ * EventSectionFinder tests.
+ *
+ * @package DataMachineEvents\Tests\Unit
+ */
+
+namespace DataMachineEvents\Tests\Unit;
+
+use DataMachine\Core\ExecutionContext;
+use DataMachine\Core\Steps\Fetch\FreshCandidateCollector;
+use DataMachineEvents\Steps\EventImport\Handlers\WebScraper\EventSectionFinder;
+use WP_UnitTestCase;
+
+class EventSectionFinderTest extends WP_UnitTestCase {
+
+	public function test_processed_first_section_does_not_stop_scan_for_later_fresh_section(): void {
+		$context = $this->createMock( ExecutionContext::class );
+		$context->method( 'isDirect' )->willReturn( true );
+		$context->expects( $this->exactly( 2 ) )
+			->method( 'isItemProcessed' )
+			->willReturnOnConsecutiveCalls( true, false );
+		$context->expects( $this->once() )
+			->method( 'isItemClaimed' )
+			->willReturn( false );
+
+		$collector = new FreshCandidateCollector( $context );
+		$finder    = new EventSectionFinder(
+			$collector,
+			fn ( string $html ): string => trim( $html ),
+			fn ( string $ymd ): bool => false
+		);
+
+		$html = '<html><body>'
+			. '<div class="event"><h2>Processed Show</h2>'
+			. '<p>This processed event section has enough content to pass the length checks.</p></div>'
+			. '<div class="event"><h2>Fresh Show</h2>'
+			. '<p>This fresh event section has enough content to pass the length checks.</p></div>'
+			. '</body></html>';
+
+		$section = $finder->find_first_eligible_section( $html, 'https://example.com/events', $context );
+
+		$this->assertNotNull( $section );
+		$this->assertStringContainsString( 'Fresh Show', $section['raw_html'] );
+		$this->assertSame(
+			array(
+				'raw_seen'           => 2,
+				'accepted'           => 1,
+				'processed_skipped'  => 1,
+				'claimed_skipped'    => 0,
+				'duplicate_skipped'  => 0,
+				'reprocess_accepted' => 0,
+				'max_items'          => 0,
+				'source_exhausted'   => false,
+			),
+			$collector->getDiagnostics()
+		);
+	}
+}