diff --git a/scrape/helpers_test.go b/scrape/helpers_test.go
index f847dbac76..ff7a7bf65a 100644
--- a/scrape/helpers_test.go
+++ b/scrape/helpers_test.go
@@ -19,7 +19,6 @@ import (
 	"encoding/binary"
 	"fmt"
 	"math"
-	"math/rand"
 	"strings"
 	"sync"
 	"testing"
@@ -148,10 +147,11 @@ func (a *collectResultAppender) Append(ref storage.SeriesRef, lset labels.Labels
 		f:      v,
 	})
 
-	if ref == 0 {
-		ref = storage.SeriesRef(rand.Uint64())
-	}
 	if a.next == nil {
+		if ref == 0 {
+			// Use labels hash as a stand-in for unique series reference, to avoid having to track all series.
+			ref = storage.SeriesRef(lset.Hash())
+		}
 		return ref, nil
 	}
 
@@ -195,10 +195,10 @@ func (a *collectResultAppender) UpdateMetadata(ref storage.SeriesRef, l labels.L
 	a.mtx.Lock()
 	defer a.mtx.Unlock()
 	a.pendingMetadata = append(a.pendingMetadata, metadataEntry{metric: l, m: m})
-	if ref == 0 {
-		ref = storage.SeriesRef(rand.Uint64())
-	}
 	if a.next == nil {
+		if ref == 0 {
+			ref = storage.SeriesRef(l.Hash())
+		}
 		return ref, nil
 	}
 
diff --git a/scrape/scrape.go b/scrape/scrape.go
index d10858d8ae..db662cb089 100644
--- a/scrape/scrape.go
+++ b/scrape/scrape.go
@@ -996,11 +996,9 @@ type scrapeCache struct {
 	// be a pointer so we can update it.
 	droppedSeries map[string]*uint64
 
-	// seriesCur and seriesPrev store the labels of series that were seen
-	// in the current and previous scrape.
-	// We hold two maps and swap them out to save allocations.
-	seriesCur  map[uint64]*cacheEntry
-	seriesPrev map[uint64]*cacheEntry
+	// Series that were seen in the current and previous scrape, for staleness detection.
+	seriesCur  map[storage.SeriesRef]*cacheEntry
+	seriesPrev map[storage.SeriesRef]*cacheEntry
 
 	// TODO(bwplotka): Consider moving Metadata API to use WAL instead of scrape loop to
 	// avoid locking (using metadata API can block scraping).
@@ -1027,8 +1025,8 @@ func newScrapeCache(metrics *scrapeMetrics) *scrapeCache {
 	return &scrapeCache{
 		series:        map[string]*cacheEntry{},
 		droppedSeries: map[string]*uint64{},
-		seriesCur:     map[uint64]*cacheEntry{},
-		seriesPrev:    map[uint64]*cacheEntry{},
+		seriesCur:     map[storage.SeriesRef]*cacheEntry{},
+		seriesPrev:    map[storage.SeriesRef]*cacheEntry{},
 		metadata:      map[string]*metaEntry{},
 		metrics:       metrics,
 	}
@@ -1076,13 +1074,9 @@ func (c *scrapeCache) iterDone(flushCache bool) {
 		c.metaMtx.Unlock()
 	}
 
-	// Swap current and previous series.
+	// Swap current and previous series then clear the new current, to save allocations.
 	c.seriesPrev, c.seriesCur = c.seriesCur, c.seriesPrev
-
-	// We have to delete every single key in the map.
-	for k := range c.seriesCur {
-		delete(c.seriesCur, k)
-	}
+	clear(c.seriesCur)
 
 	c.iter++
 }
@@ -1119,13 +1113,13 @@ func (c *scrapeCache) getDropped(met []byte) bool {
 	return ok
 }
 
-func (c *scrapeCache) trackStaleness(hash uint64, ce *cacheEntry) {
-	c.seriesCur[hash] = ce
+func (c *scrapeCache) trackStaleness(ref storage.SeriesRef, ce *cacheEntry) {
+	c.seriesCur[ref] = ce
 }
 
 func (c *scrapeCache) forEachStale(f func(storage.SeriesRef, labels.Labels) bool) {
-	for h, ce := range c.seriesPrev {
-		if _, ok := c.seriesCur[h]; !ok {
+	for ref, ce := range c.seriesPrev {
+		if _, ok := c.seriesCur[ref]; !ok {
 			if !f(ce.ref, ce.lset) {
 				break
 			}
@@ -1833,7 +1827,7 @@ loop:
 
 		if err == nil {
 			if (parsedTimestamp == nil || sl.trackTimestampsStaleness) && ce != nil {
-				sl.cache.trackStaleness(ce.hash, ce)
+				sl.cache.trackStaleness(ce.ref, ce)
 			}
 		}
 
@@ -1854,7 +1848,7 @@ loop:
 			if ce != nil && (parsedTimestamp == nil || sl.trackTimestampsStaleness) {
 				// Bypass staleness logic if there is an explicit timestamp.
 				// But make sure we only do this if we have a cache entry (ce) for our series.
-				sl.cache.trackStaleness(hash, ce)
+				sl.cache.trackStaleness(ref, ce)
 			}
 			if sampleAdded && sampleLimitErr == nil && bucketLimitErr == nil {
 				seriesAdded++