diff --git a/core/txpool/blobpool/blobpool.go b/core/txpool/blobpool/blobpool.go
index b918925bd2..2a2e1080b0 100644
--- a/core/txpool/blobpool/blobpool.go
+++ b/core/txpool/blobpool/blobpool.go
@@ -543,6 +543,8 @@ type BlobPool struct {
 	stored uint64         // Useful data size of all transactions on disk
 	limbo  *limbo         // Persistent data store for the non-finalized blobs
 
+	cQueue *conversionQueue
+
 	gapped       map[common.Address][]*BlobTxForPool // Transactions that are currently gapped (nonce too high)
 	gappedSource map[common.Hash]common.Address      // Source of gapped transactions to allow rechecking on inclusion
 
@@ -642,11 +644,14 @@ func (p *BlobPool) Init(gasTip uint64, head *types.Header, reserver txpool.Reser
 	)
 	index := func(id uint64, size uint32, blob []byte) {
 		err := p.parseTransaction(id, size, blob)
-		if err != nil {
-			toDelete = append(toDelete, id)
-		}
+		// Transactions in legacy format will be queued for cell computation.
+		// This entry will be swapped from the store after the conversion.
 		if errors.Is(err, errLegacyTx) {
 			convertTxs = append(convertTxs, id)
+			return
+		}
+		if err != nil {
+			toDelete = append(toDelete, id)
 		}
 	}
 	store, err := billy.Open(billy.Options{Path: queuedir, Repair: true}, slotter, index)
@@ -655,52 +660,7 @@ func (p *BlobPool) Init(gasTip uint64, head *types.Header, reserver txpool.Reser
 	}
 	p.store = store
 
-	// Migrate legacy transactions (types.Transaction) to pooledBlobTx format.
-	if len(convertTxs) > 0 {
-		for _, id := range convertTxs {
-			var tx types.Transaction
-			data, err := p.store.Get(id)
-			if err != nil {
-				continue
-			}
-			err = rlp.DecodeBytes(data, &tx)
-			if err != nil {
-				continue
-			}
-			if tx.BlobTxSidecar() == nil {
-				continue
-			}
-			ptx, err := newBlobTxForPool(&tx)
-			// Note that we skip errors here.
-			// Just like parseTransaction failure does not abort the blobpool creation,
-			// conversion process also cannot abort the entire process.
-			if err != nil {
-				log.Error("Failed to convert legacy tx to pooledBlobTx", "hash", tx.Hash(), "err", err)
-				continue
-			}
-			blob, err := rlp.EncodeToBytes(ptx)
-			if err != nil {
-				continue
-			}
-			id, err := p.store.Put(blob)
-			if err != nil {
-				continue
-			}
-			meta := newBlobTxMeta(id, p.store.Size(id), ptx)
-
-			// If the newly inserted transaction fails to be tracked,
-			// it should also be removed with those in `toDelete`
-			sender, err := types.Sender(p.signer, ptx.Tx)
-			if err != nil {
-				toDelete = append(toDelete, id)
-				continue
-			}
-			if err := p.trackTransaction(meta, sender); err != nil {
-				toDelete = append(toDelete, id)
-				continue
-			}
-		}
-	}
+	p.cQueue = newConversionQueue()
 
 	if len(toDelete) > 0 {
 		log.Warn("Dropping invalidated blob transactions", "ids", toDelete)
@@ -744,7 +704,8 @@ func (p *BlobPool) Init(gasTip uint64, head *types.Header, reserver txpool.Reser
 
 	// Pool initialized, attach the blob limbo to it to track blobs included
 	// recently but not yet finalized
-	p.limbo, err = newLimbo(p.chain.Config(), limbodir)
+	var convertLimbo []uint64
+	p.limbo, convertLimbo, err = newLimbo(p.chain.Config(), limbodir)
 	if err != nil {
 		p.Close()
 		return err
@@ -763,12 +724,22 @@ func (p *BlobPool) Init(gasTip uint64, head *types.Header, reserver txpool.Reser
 	// Update the metrics and return the constructed pool
 	datacapGauge.Update(int64(p.config.Datacap))
 	p.updateStorageMetrics()
+
+	if len(convertTxs) > 0 {
+		p.cQueue.launchConversion(func() { p.convertLegacyTxs(convertTxs) })
+	}
+	if len(convertLimbo) > 0 {
+		p.cQueue.launchConversion(func() { p.convertLegacyLimbo(convertLimbo) })
+	}
 	return nil
 }
 
 // Close closes down the underlying persistent store.
 func (p *BlobPool) Close() error {
 	var errs []error
+	if p.cQueue != nil {
+		p.cQueue.close()
+	}
 	if p.limbo != nil { // Close might be invoked due to error in constructor, before p,limbo is set
 		if err := p.limbo.Close(); err != nil {
 			errs = append(errs, err)
@@ -817,6 +788,91 @@ func (p *BlobPool) parseTransaction(id uint64, size uint32, blob []byte) error {
 	return p.trackTransaction(meta, sender)
 }
 
+func (p *BlobPool) convertLegacyTxs(ids []uint64) {
+	start := time.Now()
+	var converted, discarded int
+
+	for _, id := range ids {
+		p.lock.Lock()
+		data, err := p.store.Get(id)
+		if err != nil {
+			p.lock.Unlock()
+			continue
+		}
+		if derr := p.store.Delete(id); derr != nil {
+			log.Error("Failed to delete legacy blob tx", "id", id, "err", derr)
+			p.lock.Unlock()
+			continue
+		}
+		p.lock.Unlock()
+		var tx types.Transaction
+		if err := rlp.DecodeBytes(data, &tx); err != nil {
+			log.Error("Failed to decode legacy blob tx", "id", id, "err", err)
+			continue
+		}
+		ptx, err := newBlobTxForPool(&tx)
+		if err != nil {
+			log.Error("Failed to convert legacy blob tx", "hash", tx.Hash(), "err", err)
+			continue
+		}
+		if err := p.AddPooledTx(ptx); err != nil {
+			log.Debug("Discarded converted blob tx", "hash", tx.Hash(), "err", err)
+			discarded++
+		} else {
+			converted++
+		}
+	}
+
+	log.Info("Completed blob transaction conversion", "converted", converted, "discarded", discarded, "elapsed", common.PrettyDuration(time.Since(start)))
+}
+
+func (p *BlobPool) convertLegacyLimbo(ids []uint64) {
+	start := time.Now()
+	var converted, discarded int
+
+	for _, id := range ids {
+		p.lock.Lock()
+		data, err := p.limbo.store.Get(id)
+		if err != nil {
+			p.lock.Unlock()
+			continue
+		}
+		if derr := p.limbo.store.Delete(id); derr != nil {
+			log.Error("Failed to delete legacy blob tx", "id", id, "err", derr)
+			p.lock.Unlock()
+			continue
+		}
+		p.lock.Unlock()
+		var legacy struct {
+			TxHash common.Hash
+			Block  uint64
+			Tx     *types.Transaction
+		}
+		if err := rlp.DecodeBytes(data, &legacy); err != nil {
+			log.Error("Failed to decode legacy limbo entry", "id", id, "err", err)
+			continue
+		}
+		if legacy.Tx == nil || legacy.Tx.BlobTxSidecar() == nil {
+			continue
+		}
+		ptx, err := newBlobTxForPool(legacy.Tx)
+		if err != nil {
+			log.Error("Failed to convert legacy limbo entry", "hash", legacy.TxHash, "err", err)
+			continue
+		}
+		p.lock.Lock()
+		if err := p.limbo.setAndIndex(ptx, legacy.Block); err != nil {
+			log.Error("Failed to re-store converted limbo entry", "hash", legacy.TxHash, "err", err)
+			discarded++
+		} else {
+			converted++
+		}
+		p.lock.Unlock()
+	}
+
+	log.Info("Completed limbo blob conversion", "converted", converted, "discarded", discarded, "elapsed", common.PrettyDuration(time.Since(start)))
+}
+
 // trackTransaction registers a transaction's metadata in the pool's indices.
 func (p *BlobPool) trackTransaction(meta *blobTxMeta, sender common.Address) error {
 	if p.lookup.exists(meta.hash) {
@@ -1904,7 +1960,7 @@ func (p *BlobPool) Add(txs []*types.Transaction, sync bool) []error {
 		if errs[i] = p.ValidateTxBasics(tx); errs[i] != nil {
 			continue
 		}
-		ptx, err := newBlobTxForPool(tx)
+		ptx, err := p.cQueue.convert(tx)
 		if err != nil {
 			errs[i] = err
 			continue
diff --git a/core/txpool/blobpool/blobpool_test.go b/core/txpool/blobpool/blobpool_test.go
index 394b9e31d4..4e5f4f5e5a 100644
--- a/core/txpool/blobpool/blobpool_test.go
+++ b/core/txpool/blobpool/blobpool_test.go
@@ -31,6 +31,7 @@ import (
 	"slices"
 	"sync"
 	"testing"
+	"time"
 
 	"github.com/ethereum/go-ethereum/common"
 	"github.com/ethereum/go-ethereum/consensus/misc/eip1559"
@@ -1334,6 +1335,14 @@ func TestLegacyTxConversion(t *testing.T) {
 	}
 	defer pool.Close()
 
+	deadline := time.Now().Add(30 * time.Second)
+	for pool.Get(tx1.Hash()) == nil || pool.Get(tx2.Hash()) == nil {
+		if time.Now().After(deadline) {
+			t.Fatalf("legacy txs were not converted in time")
+		}
+		time.Sleep(10 * time.Millisecond)
+	}
+
 	// Both transactions should be retrievable.
 	for _, want := range []*types.Transaction{tx1, tx2} {
 		got := pool.Get(want.Hash())
@@ -1359,6 +1368,75 @@ func TestLegacyTxConversion(t *testing.T) {
 	verifyPoolInternals(t, pool)
 }
 
+func TestLegacyLimboConversion(t *testing.T) {
+	storage := t.TempDir()
+	os.MkdirAll(filepath.Join(storage, pendingTransactionStore), 0700)
+	limbodir := filepath.Join(storage, limboedTransactionStore)
+	os.MkdirAll(limbodir, 0700)
+
+	key, _ := crypto.GenerateKey()
+	tx := makeMultiBlobTx(0, 1, 1000, 100, 2, 0, key)
+
+	store, err := billy.Open(billy.Options{Path: limbodir}, newSlotterEIP7594(testMaxBlobsPerBlock), nil)
+	if err != nil {
+		t.Fatalf("failed to open limbo billy: %v", err)
+	}
+	data, err := rlp.EncodeToBytes(&struct {
+		TxHash common.Hash
+		Block  uint64
+		Tx     *types.Transaction
+	}{TxHash: tx.Hash(), Block: 42, Tx: tx})
+	if err != nil {
+		t.Fatalf("failed to legacy-encode limbo entry: %v", err)
+	}
+	if _, err := store.Put(data); err != nil {
+		t.Fatalf("failed to put legacy limbo entry: %v", err)
+	}
+	store.Close()
+
+	statedb, _ := state.New(types.EmptyRootHash, state.NewDatabaseForTesting())
+	statedb.Commit(0, true, false)
+	chain := &testBlockChain{
+		config:  params.MainnetChainConfig,
+		basefee: uint256.NewInt(params.InitialBaseFee),
+		blobfee: uint256.NewInt(params.BlobTxMinBlobGasprice),
+		statedb: statedb,
+	}
+	pool := New(Config{Datadir: storage}, chain, nil)
+	if err := pool.Init(1, chain.CurrentBlock(), newReserver()); err != nil {
+		t.Fatalf("failed to create blob pool: %v", err)
+	}
+	defer pool.Close()
+
+	deadline := time.Now().Add(30 * time.Second)
+	for {
+		pool.lock.RLock()
+		_, ok := pool.limbo.index[tx.Hash()]
+		pool.lock.RUnlock()
+		if ok {
+			break
+		}
+		if time.Now().After(deadline) {
+			t.Fatalf("legacy limbo entry was not converted in time")
+		}
+		time.Sleep(10 * time.Millisecond)
+	}
+
+	pool.lock.Lock()
+	ptx, err := pool.limbo.pull(tx.Hash())
+	pool.lock.Unlock()
+	if err != nil {
+		t.Fatalf("failed to pull converted limbo entry: %v", err)
+	}
+	full, err := ptx.toTx()
+	if err != nil {
+		t.Fatalf("failed to reconstruct tx from converted limbo entry: %v", err)
+	}
+	if full.Hash() != tx.Hash() {
+		t.Fatalf("converted limbo tx hash mismatch: have %s, want %s", full.Hash(), tx.Hash())
+	}
+}
+
 // TestBlobCountLimit tests the blobpool enforced limits on the max blob count.
 func TestBlobCountLimit(t *testing.T) {
 	var (
diff --git a/core/txpool/blobpool/conversion.go b/core/txpool/blobpool/conversion.go
new file mode 100644
index 0000000000..714975cce3
--- /dev/null
+++ b/core/txpool/blobpool/conversion.go
@@ -0,0 +1,207 @@
+// Copyright 2026 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package blobpool
+
+import (
+	"errors"
+	"slices"
+	"sync/atomic"
+	"time"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/log"
+)
+
+// maxPendingConversionTasks caps the number of pending conversion tasks. This
+// prevents excessive memory usage; the worst-case scenario (2k transactions
+// with 6 blobs each) would consume approximately 1.5GB of memory.
+const maxPendingConversionTasks = 2048
+
+type convertResult struct {
+	ptx *BlobTxForPool
+	err error
+}
+
+// txConvert represents a conversion task with an attached legacy blob transaction.
+type txConvert struct {
+	tx   *types.Transaction // Legacy blob transaction
+	done chan convertResult // Channel for signaling back if the conversion succeeds
+}
+
+// conversionQueue is a dedicated queue for converting legacy blob transactions
+// received from the network after the Osaka fork. Since conversion is expensive,
+// it is performed in the background by a single thread, ensuring the main Geth
+// process is not overloaded.
+type conversionQueue struct {
+	tasks           chan *txConvert
+	startConversion chan func()
+	quit            chan struct{}
+	closed          chan struct{}
+
+	queue    []func()
+	taskDone chan struct{}
+}
+
+// newConversionQueue constructs the conversion queue.
+func newConversionQueue() *conversionQueue {
+	q := &conversionQueue{
+		tasks:           make(chan *txConvert),
+		startConversion: make(chan func()),
+		quit:            make(chan struct{}),
+		closed:          make(chan struct{}),
+	}
+	go q.loop()
+	return q
+}
+
+// convert accepts a legacy blob transaction with version-0 blobs and queues it
+// for conversion.
+//
+// This function may block for a long time until the transaction is processed.
+func (q *conversionQueue) convert(tx *types.Transaction) (*BlobTxForPool, error) {
+	done := make(chan convertResult, 1)
+	select {
+	case q.tasks <- &txConvert{tx: tx, done: done}:
+		res := <-done
+		return res.ptx, res.err
+	case <-q.closed:
+		return nil, errors.New("conversion queue closed")
+	}
+}
+
+// launchConversion starts a conversion task in the background.
+func (q *conversionQueue) launchConversion(fn func()) error {
+	select {
+	case q.startConversion <- fn:
+		return nil
+	case <-q.closed:
+		return errors.New("conversion queue closed")
+	}
+}
+
+// close terminates the conversion queue.
+func (q *conversionQueue) close() {
+	select {
+	case <-q.closed:
+		return
+	default:
+		close(q.quit)
+		<-q.closed
+	}
+}
+
+// run converts a batch of legacy blob txs to the new cell proof format.
+func (q *conversionQueue) run(tasks []*txConvert, done chan struct{}, interrupt *atomic.Int32) {
+	defer close(done)
+
+	for _, t := range tasks {
+		if interrupt != nil && interrupt.Load() != 0 {
+			t.done <- convertResult{err: errors.New("conversion is interrupted")}
+			continue
+		}
+		// Run the conversion, the original sidecar will be mutated in place
+		start := time.Now()
+		ptx, err := newBlobTxForPool(t.tx)
+		t.done <- convertResult{ptx: ptx, err: err}
+		log.Trace("Converted legacy blob tx", "hash", t.tx.Hash(), "err", err, "elapsed", common.PrettyDuration(time.Since(start)))
+	}
+}
+
+func (q *conversionQueue) loop() {
+	defer close(q.closed)
+
+	var (
+		done      chan struct{} // Non-nil if background routine is active
+		interrupt *atomic.Int32 // Flag to signal conversion interruption
+
+		// The pending tasks for sidecar conversion. We assume the number of legacy
+		// blob transactions requiring conversion will not be excessive. However,
+		// a hard cap is applied as a protective measure.
+		txTasks []*txConvert
+	)
+
+	for {
+		select {
+		case t := <-q.tasks:
+			if len(txTasks) >= maxPendingConversionTasks {
+				t.done <- convertResult{err: errors.New("conversion queue is overloaded")}
+				continue
+			}
+			txTasks = append(txTasks, t)
+
+			// Launch the background conversion thread if it's idle
+			if done == nil {
+				done, interrupt = make(chan struct{}), new(atomic.Int32)
+
+				tasks := slices.Clone(txTasks)
+				txTasks = txTasks[:0]
+				go q.run(tasks, done, interrupt)
+			}
+
+		case <-done:
+			done, interrupt = nil, nil
+			if len(txTasks) > 0 {
+				done, interrupt = make(chan struct{}), new(atomic.Int32)
+				tasks := slices.Clone(txTasks)
+				txTasks = txTasks[:0]
+				go q.run(tasks, done, interrupt)
+			}
+
+		case fn := <-q.startConversion:
+			q.queue = append(q.queue, fn)
+			q.runNextTask()
+
+		case <-q.taskDone:
+			q.runNextTask()
+
+		case <-q.quit:
+			if done != nil {
+				log.Debug("Waiting for blob proof conversion to exit")
+				interrupt.Store(1)
+				<-done
+			}
+			if q.taskDone != nil {
+				log.Debug("Waiting for blobpool billy conversion to exit")
+				<-q.taskDone
+			}
+			// Signal any tasks that were queued for the next batch but never started
+			// so callers blocked in convert() receive an error instead of hanging.
+			for _, t := range txTasks {
+				// Best-effort notify; t.done is a buffered channel of size 1
+				// created by convert(), and we send exactly once per task.
+				t.done <- convertResult{err: errors.New("conversion queue closed")}
+			}
+			// Drop references to allow GC of the backing array.
+			txTasks = txTasks[:0]
+			return
+		}
+	}
+}
+
+func (q *conversionQueue) runNextTask() {
+	if len(q.queue) == 0 {
+		q.taskDone = nil
+		return
+	}
+	fn := q.queue[0]
+	q.queue = append(q.queue[:0], q.queue[1:]...)
+
+	done := make(chan struct{})
+	go func() { defer close(done); fn() }()
+	q.taskDone = done
+}
diff --git a/core/txpool/blobpool/conversion_test.go b/core/txpool/blobpool/conversion_test.go
new file mode 100644
index 0000000000..b52239641e
--- /dev/null
+++ b/core/txpool/blobpool/conversion_test.go
@@ -0,0 +1,114 @@
+// Copyright 2026 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package blobpool
+
+import (
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/ethereum/go-ethereum/crypto"
+	"github.com/ethereum/go-ethereum/params"
+)
+
+func TestConversionQueueBasic(t *testing.T) {
+	queue := newConversionQueue()
+	defer queue.close()
+
+	key, _ := crypto.GenerateKey()
+	tx := makeTx(0, 1, 1, 1, key)
+
+	ptx, err := queue.convert(tx)
+	if err != nil {
+		t.Fatalf("Expected successful conversion, got error: %v", err)
+	}
+	if ptx == nil {
+		t.Fatal("Expected a converted transaction, got nil")
+	}
+	if ptx.Tx.Hash() != tx.Hash() {
+		t.Errorf("Converted tx hash mismatch: have %s, want %s", ptx.Tx.Hash(), tx.Hash())
+	}
+	if len(ptx.CellSidecar.Cells) == 0 {
+		t.Error("Expected cells to be computed during conversion")
+	}
+}
+
+func TestConversionQueueClosed(t *testing.T) {
+	queue := newConversionQueue()
+	queue.close()
+
+	key, _ := crypto.GenerateKey()
+	tx := makeTx(0, 1, 1, 1, key)
+
+	if _, err := queue.convert(tx); err == nil {
+		t.Fatal("Expected error when converting on closed queue, got nil")
+	}
+}
+
+func TestConversionQueueDoubleClose(t *testing.T) {
+	queue := newConversionQueue()
+	queue.close()
+	queue.close() // Should not panic
+}
+
+func TestConversionQueueAutoRestartBatch(t *testing.T) {
+	queue := newConversionQueue()
+	defer queue.close()
+
+	key, _ := crypto.GenerateKey()
+
+	// Create a heavy transaction to ensure the first batch runs long enough
+	// for subsequent tasks to be queued while it is active.
+	heavy := makeMultiBlobTx(0, 1, 1, 1, int(params.BlobTxMaxBlobs), 0, key)
+
+	var wg sync.WaitGroup
+	wg.Add(1)
+	heavyDone := make(chan error, 1)
+	go func() {
+		defer wg.Done()
+		_, err := queue.convert(heavy)
+		heavyDone <- err
+	}()
+
+	// Give the conversion worker a head start so that the following tasks are
+	// enqueued while the first batch is running.
+	time.Sleep(200 * time.Millisecond)
+
+	tx1 := makeTx(1, 1, 1, 1, key)
+	tx2 := makeTx(2, 1, 1, 1, key)
+
+	wg.Add(2)
+	done1 := make(chan error, 1)
+	done2 := make(chan error, 1)
+	go func() { defer wg.Done(); _, err := queue.convert(tx1); done1 <- err }()
+	go func() { defer wg.Done(); _, err := queue.convert(tx2); done2 <- err }()
+
+	for _, c := range []struct {
+		name string
+		done chan error
+	}{{"tx1", done1}, {"tx2", done2}, {"heavy", heavyDone}} {
+		select {
+		case err := <-c.done:
+			if err != nil {
+				t.Fatalf("%s conversion error: %v", c.name, err)
+			}
+		case <-time.After(30 * time.Second):
+			t.Fatalf("timeout waiting for %s conversion", c.name)
+		}
+	}
+	wg.Wait()
+}
diff --git a/core/txpool/blobpool/limbo.go b/core/txpool/blobpool/limbo.go
index dc091132bb..8e44f51fdd 100644
--- a/core/txpool/blobpool/limbo.go
+++ b/core/txpool/blobpool/limbo.go
@@ -49,7 +49,7 @@ type limbo struct {
 }
 
 // newLimbo opens and indexes a set of limboed blob transactions.
-func newLimbo(config *params.ChainConfig, datadir string) (*limbo, error) {
+func newLimbo(config *params.ChainConfig, datadir string) (*limbo, []uint64, error) {
 	l := &limbo{
 		index:  make(map[common.Hash]uint64),
 		groups: make(map[uint64]map[uint64]common.Hash),
@@ -61,19 +61,27 @@ func newLimbo(config *params.ChainConfig, datadir string) (*limbo, error) {
 	// See if we need to migrate the limbo after fusaka.
 	slotter, err := tryMigrate(config, slotter, datadir)
 	if err != nil {
-		return nil, err
+		return nil, nil, err
 	}
 
 	// Index all limboed blobs on disk and delete anything unprocessable
-	var fails []uint64
+	var (
+		fails   []uint64
+		convert []uint64
+	)
 	index := func(id uint64, size uint32, data []byte) {
-		if l.parseBlob(id, data) != nil {
+		err := l.parseBlob(id, data)
+		if errors.Is(err, errLegacyTx) {
+			convert = append(convert, id)
+			return
+		}
+		if err != nil {
 			fails = append(fails, id)
 		}
 	}
 	store, err := billy.Open(billy.Options{Path: datadir, Repair: true}, slotter, index)
 	if err != nil {
-		return nil, err
+		return nil, nil, err
 	}
 	l.store = store
 
@@ -82,11 +90,11 @@ func newLimbo(config *params.ChainConfig, datadir string) (*limbo, error) {
 		for _, id := range fails {
 			if err := l.store.Delete(id); err != nil {
 				l.Close()
-				return nil, err
+				return nil, nil, err
 			}
 		}
 	}
-	return l, nil
+	return l, convert, nil
 }
 
 // Close closes down the underlying persistent store.
@@ -99,6 +107,9 @@ func (l *limbo) Close() error {
 func (l *limbo) parseBlob(id uint64, data []byte) error {
 	item := new(limboBlob)
 	if err := rlp.DecodeBytes(data, item); err != nil {
+		if isLegacyLimboBlob(data) {
+			return errLegacyTx
+		}
 		// This path is impossible unless the disk data representation changes
 		// across restarts. For that ever improbable case, recover gracefully
 		// by ignoring this data entry.
@@ -122,6 +133,17 @@ func (l *limbo) parseBlob(id uint64, data []byte) error {
 	return nil
 }
 
+// isLegacyLimboBlob returns true if the data is encoded in legacy limboBlob type.
+// It checks whether the first byte of third element is blobTxType.
+func isLegacyLimboBlob(data []byte) bool {
+	elems, err := rlp.SplitListValues(data)
+	if err != nil || len(elems) < 3 {
+		return false
+	}
+	kind, content, _, err := rlp.Split(elems[2])
+	return err == nil && kind == rlp.String && len(content) > 1 && content[0] == types.BlobTxType
+}
+
 // finalize evicts all blobs belonging to a recently finalized block or older.
 func (l *limbo) finalize(final *types.Header) {
 	// Just in case there's no final block yet (network not yet merged, weird