diff --git a/internal/federation/federatingdb/db.go b/internal/federation/federatingdb/db.go
index f6587a1b..5f8b9ad9 100644
--- a/internal/federation/federatingdb/db.go
+++ b/internal/federation/federatingdb/db.go
@@ -21,6 +21,7 @@ package federatingdb
 import (
 	"context"
 	"sync"
+	"time"
 
 	"github.com/go-fed/activity/pub"
 	"github.com/go-fed/activity/streams/vocab"
@@ -41,7 +42,9 @@ type DB interface {
 // FederatingDB uses the underlying DB interface to implement the go-fed pub.Database interface.
 // It doesn't care what the underlying implementation of the DB interface is, as long as it works.
 type federatingDB struct {
-	locks         *sync.Map
+	mutex         sync.Mutex
+	locks         map[string]*mutex
+	pool          sync.Pool
 	db            db.DB
 	config        *config.Config
 	log           *logrus.Logger
@@ -50,11 +53,32 @@ type federatingDB struct {
 
 // New returns a DB interface using the given database, config, and logger.
 func New(db db.DB, config *config.Config, log *logrus.Logger) DB {
-	return &federatingDB{
-		locks:         new(sync.Map),
+	fdb := federatingDB{
+		mutex:         sync.Mutex{},
+		locks:         make(map[string]*mutex, 100),
+		pool:          sync.Pool{New: func() interface{} { return &mutex{} }},
 		db:            db,
 		config:        config,
 		log:           log,
 		typeConverter: typeutils.NewConverter(config, db),
 	}
+	go fdb.cleanupLocks()
+	return &fdb
+}
+
+func (db *federatingDB) cleanupLocks() {
+	for {
+		// Sleep for a minute...
+		time.Sleep(time.Minute)
+
+		// Delete unused locks from map
+		db.mutex.Lock()
+		for id, mu := range db.locks {
+			if !mu.inUse() {
+				delete(db.locks, id)
+				db.pool.Put(mu)
+			}
+		}
+		db.mutex.Unlock()
+	}
 }
diff --git a/internal/federation/federatingdb/lock.go b/internal/federation/federatingdb/lock.go
index c9062da8..0d35f337 100644
--- a/internal/federation/federatingdb/lock.go
+++ b/internal/federation/federatingdb/lock.go
@@ -23,6 +23,7 @@ import (
 	"errors"
 	"net/url"
 	"sync"
+	"sync/atomic"
 )
 
 // Lock takes a lock for the object at the specified id. If an error
@@ -45,14 +46,21 @@ func (f *federatingDB) Lock(c context.Context, id *url.URL) error {
 	if id == nil {
 		return errors.New("Lock: id was nil")
 	}
+	idStr := id.String()
 
-	mu := &sync.Mutex{}
-	mu.Lock() // Optimistically lock if we do store it.
-	i, loaded := f.locks.LoadOrStore(id.String(), mu)
-	if loaded {
-		mu = i.(*sync.Mutex)
-		mu.Lock()
+	// Acquire map lock
+	f.mutex.Lock()
+
+	// Get mutex, or create new
+	mu, ok := f.locks[idStr]
+	if !ok {
+		mu = f.pool.Get().(*mutex)
+		f.locks[idStr] = mu
 	}
+
+	// Unlock map, acquire mutex lock
+	f.mutex.Unlock()
+	mu.Lock()
 	return nil
 }
 
@@ -66,12 +74,43 @@ func (f *federatingDB) Unlock(c context.Context, id *url.URL) error {
 	if id == nil {
 		return errors.New("Unlock: id was nil")
 	}
+	idStr := id.String()
+
+	// Check map for mutex
+	f.mutex.Lock()
+	mu, ok := f.locks[idStr]
+	f.mutex.Unlock()
 
-	i, ok := f.locks.Load(id.String())
 	if !ok {
 		return errors.New("missing an id in unlock")
 	}
-	mu := i.(*sync.Mutex)
+
+	// Unlock the mutex
 	mu.Unlock()
 	return nil
 }
+
+// mutex defines a mutex we can check the lock status of.
+// this is not perfect, but it's good enough for a semi
+// regular mutex cleanup routine
+type mutex struct {
+	mu sync.Mutex
+	st uint32
+}
+
+// inUse returns if the mutex is in use
+func (mu *mutex) inUse() bool {
+	return atomic.LoadUint32(&mu.st) == 1
+}
+
+// Lock acquire mutex lock
+func (mu *mutex) Lock() {
+	mu.mu.Lock()
+	atomic.StoreUint32(&mu.st, 1)
+}
+
+// Unlock releases mutex lock
+func (mu *mutex) Unlock() {
+	mu.mu.Unlock()
+	atomic.StoreUint32(&mu.st, 0)
+}