diff --git a/spvsvc/spvchain/blockmanager.go b/spvsvc/spvchain/blockmanager.go
index 307f63f..e9612e2 100644
--- a/spvsvc/spvchain/blockmanager.go
+++ b/spvsvc/spvchain/blockmanager.go
@@ -162,8 +162,10 @@ type blockManager struct {
 
 	basicHeaders            map[chainhash.Hash]map[chainhash.Hash][]*serverPeer
 	lastBasicCFHeaderHeight int32
+	numBasicCFHeadersMsgs   int32
 	extendedHeaders         map[chainhash.Hash]map[chainhash.Hash][]*serverPeer
 	lastExtCFHeaderHeight   int32
+	numExtCFHeadersMsgs     int32
 	mapMutex                sync.Mutex
 
 	minRetargetTimespan int64 // target timespan / adjustment factor
@@ -1067,6 +1069,13 @@ func (b *blockManager) QueueCFHeaders(cfheaders *wire.MsgCFHeaders,
 		return
 	}
 
+	// Track number of pending cfheaders messsages for both basic and
+	// extended filters.
+	pendingMsgs := &b.numBasicCFHeadersMsgs
+	if cfheaders.Extended {
+		pendingMsgs = &b.numExtCFHeadersMsgs
+	}
+	atomic.AddInt32(pendingMsgs, 1)
 	b.peerChan <- &cfheadersMsg{cfheaders: cfheaders, peer: sp}
 }
 
@@ -1079,7 +1088,14 @@ func (b *blockManager) handleCFHeadersMsg(cfhmsg *cfheadersMsg) {
 		extended: cfhmsg.cfheaders.Extended,
 		stopHash: cfhmsg.cfheaders.StopHash,
 	}
+	headerMap := b.basicHeaders
+	pendingMsgs := &b.numBasicCFHeadersMsgs
+	if req.extended {
+		headerMap = b.extendedHeaders
+		pendingMsgs = &b.numExtCFHeadersMsgs
+	}
 	defer delete(cfhmsg.peer.requestedCFHeaders, req)
+	defer atomic.AddInt32(pendingMsgs, -1)
 	// Check that the count is correct. This works even when the map lookup
 	// fails as it returns 0 in that case.
 	headerList := cfhmsg.cfheaders.HeaderHashes
@@ -1106,10 +1122,6 @@ func (b *blockManager) handleCFHeadersMsg(cfhmsg *cfheadersMsg) {
 		return
 	}
 	// Cycle through the filter header hashes and process them.
-	filterMap := b.basicHeaders
-	if req.extended {
-		filterMap = b.extendedHeaders
-	}
 	var node *headerNode
 	var hash chainhash.Hash
 	for i := respLen - 1; i >= 0 && el != nil; i-- {
@@ -1118,13 +1130,13 @@ func (b *blockManager) handleCFHeadersMsg(cfhmsg *cfheadersMsg) {
 		// the database. Either way, break processing.
 		node = el.Value.(*headerNode)
 		hash = node.header.BlockHash()
-		if _, ok := filterMap[hash]; !ok {
+		b.mapMutex.Lock()
+		if _, ok := headerMap[hash]; !ok {
 			break
 		}
 		// Process this header and set up the next iteration.
-		b.mapMutex.Lock()
-		filterMap[hash][*headerList[i]] = append(
-			filterMap[hash][*headerList[i]], cfhmsg.peer,
+		headerMap[hash][*headerList[i]] = append(
+			headerMap[hash][*headerList[i]], cfhmsg.peer,
 		)
 		b.mapMutex.Unlock()
 		el = el.Prev()
@@ -1147,30 +1159,30 @@ func (b *blockManager) handleProcessCFHeadersMsg(msg *processCFHeadersMsg) {
 	// Assume we aren't ready to make a decision about correct headers yet.
 	ready := false
 
-	filterMap := b.basicHeaders
+	headerMap := b.basicHeaders
 	writeFunc := b.server.putBasicHeader
 	readFunc := b.server.GetBasicHeader
 	lastCFHeaderHeight := &b.lastBasicCFHeaderHeight
+	pendingMsgs := &b.numBasicCFHeadersMsgs
 	if msg.extended {
-		filterMap = b.extendedHeaders
+		headerMap = b.extendedHeaders
 		writeFunc = b.server.putExtHeader
 		readFunc = b.server.GetExtHeader
 		lastCFHeaderHeight = &b.lastExtCFHeaderHeight
+		pendingMsgs = &b.numExtCFHeadersMsgs
 	}
 
 	// If we have started receiving cfheaders messages for blocks farther
 	// than the last set we haven't made a decision on, it's time to make
 	// a decision.
-	if msg.earliestNode.height > *lastCFHeaderHeight {
-		if *lastCFHeaderHeight != 0 {
-			ready = true
-		}
-		*lastCFHeaderHeight = msg.earliestNode.height
+	if msg.earliestNode.height > *lastCFHeaderHeight+1 {
+		ready = true
 	}
 
-	// If there are no other messages left, we should go ahead and make a
-	// decision because we have all the info we're going to get.
-	if len(b.peerChan) == 0 {
+	// If there are no other cfheaders messages left for this type (basic vs
+	// extended), we should go ahead and make a decision because we have all
+	// the info we're going to get.
+	if atomic.LoadInt32(pendingMsgs) == 0 {
 		ready = true
 	}
 
@@ -1184,13 +1196,15 @@ func (b *blockManager) handleProcessCFHeadersMsg(msg *processCFHeadersMsg) {
 	// iterate through all of those headers, looking for conflicts. If we
 	// find a conflict, we have to do additional checks; otherwise, we write
 	// the filter header to the database.
+	log.Tracef("Begin processing cfheaders messages starting at %d (%s)",
+		msg.earliestNode.height, msg.earliestNode.header.BlockHash())
 	el := b.headerList.Front()
 	for el != nil {
 		node := el.Value.(*headerNode)
 		hash := node.header.BlockHash()
-		if node.height >= msg.earliestNode.height {
+		if node.height > *lastCFHeaderHeight {
 			b.mapMutex.Lock()
-			blockMap := filterMap[hash]
+			blockMap := headerMap[hash]
 			switch len(blockMap) {
 			// This should only happen if the filter has already
 			// been written to the database or if there's a reorg.
@@ -1207,13 +1221,22 @@ func (b *blockManager) handleProcessCFHeadersMsg(msg *processCFHeadersMsg) {
 			// bamboozle us (or ALL our peers are).
 			case 1:
 				// This will only cycle once
-				for filterHash := range blockMap {
-					writeFunc(hash, filterHash)
+				for headerHash := range blockMap {
+					writeFunc(hash, headerHash)
+					log.Tracef("Wrote header for block %d "+
+						"with %d cfheaders messages, "+
+						"extended: %t", node.height,
+						len(blockMap[headerHash]),
+						msg.extended)
 				}
+				*lastCFHeaderHeight = node.height
 			// This is when we have conflicting information from
 			// multiple peers.
 			// TODO: Handle this case.
 			default:
+				log.Warnf("Got more than 1 possible filter "+
+					"header for block %d (%s)", node.height,
+					node.header.BlockHash())
 			}
 			b.mapMutex.Unlock()
 		}