fetch: requests got lost when there are no peers (spacemeshos#4849)

requests are not restored back to the queue when there were no peers to dispatch to.

- restore requests when there are no peers
- push mesh to still make progress even when there are missing blocks
- increase frequency of state sync

i manually test against mainnet, artificially causing GetPeers() to return zero peers.
mesh was making very slow progress because it aborts as soon as it finds a missing block, not executing the next block even when it's available.

mesh/mesh.go

@@ -319,12 +319,15 @@ func (msh *Mesh) ProcessLayer(ctx context.Context, lid types.LayerID) error {
 			})),
 		)
 	}
-	if missing := missingBlocks(results); len(missing) > 0 {
+	applicable, missing := filterMissing(results, next)
+	if len(missing) > 0 {
 		select {
 		case <-ctx.Done():
 		case msh.missingBlocks <- missing:
 		}
-		return fmt.Errorf("request missing blocks %v", missing)
+		if len(applicable) == 0 {
+			return fmt.Errorf("request missing blocks %v", missing)
+		}
 	}
 
 	if err := msh.ensureStateConsistent(ctx, applicable); err != nil {
@@ -334,8 +337,7 @@ func (msh *Mesh) ProcessLayer(ctx context.Context, lid types.LayerID) error {
 		return err
 	}
 	if len(missing) > 0 {
-		msh.pendingUpdates.min = applicable[len(applicable)-1].Layer
-		msh.pendingUpdates.max = types.MaxLayer(msh.pendingUpdates.min, msh.pendingUpdates.max)
+		msh.pendingUpdates.min = msh.LatestLayerInState()
 	} else {
 		msh.pendingUpdates.min = 0
 		msh.pendingUpdates.max = 0