diff --git a/analyzer/consensus/consensus.go b/analyzer/consensus/consensus.go
index 2d3e1e28e..58d267efb 100644
--- a/analyzer/consensus/consensus.go
+++ b/analyzer/consensus/consensus.go
@@ -163,6 +163,9 @@ func (m *Main) Start() {
 		backoff.Success()
 		height++
 	}
+	m.logger.Info(
+		fmt.Sprintf("finished processing all blocks in the configured range [%d, %d]",
+			m.cfg.Range.From, m.cfg.Range.To))
 }
 
 // Name returns the name of the Main.
@@ -285,7 +288,7 @@ func (m *Main) processBlock(ctx context.Context, height int64) error {
 		m.queueRuntimeRegistrations,
 		m.queueEntityEvents,
 		m.queueNodeEvents,
-		m.queueRegistryEvents,
+		m.queueRegistryEventInserts,
 	} {
 		if err := f(batch, data.RegistryData); err != nil {
 			return err
@@ -293,11 +296,12 @@ func (m *Main) processBlock(ctx context.Context, height int64) error {
 	}
 
 	for _, f := range []func(*storage.QueryBatch, *storage.StakingData) error{
-		m.queueTransfers,
+		m.queueRegularTransfers,
 		m.queueBurns,
 		m.queueEscrows,
 		m.queueAllowanceChanges,
-		m.queueStakingEvents,
+		m.queueStakingEventInserts,
+		m.queueDisbursementTransfers,
 	} {
 		if err := f(batch, data.StakingData); err != nil {
 			return err
@@ -318,14 +322,14 @@ func (m *Main) processBlock(ctx context.Context, height int64) error {
 		m.queueExecutions,
 		m.queueFinalizations,
 		m.queueVotes,
-		m.queueGovernanceEvents,
+		m.queueGovernanceEventInserts,
 	} {
 		if err := f(batch, data.GovernanceData); err != nil {
 			return err
 		}
 	}
 
-	if err := m.queueRootHashEvents(batch, data.RootHashData); err != nil {
+	if err := m.queueRootHashEventInserts(batch, data.RootHashData); err != nil {
 		return err
 	}
 
@@ -566,7 +570,7 @@ func (m *Main) queueNodeEvents(batch *storage.QueryBatch, data *storage.Registry
 	return nil
 }
 
-func (m *Main) queueRegistryEvents(batch *storage.QueryBatch, data *storage.RegistryData) error {
+func (m *Main) queueRegistryEventInserts(batch *storage.QueryBatch, data *storage.RegistryData) error {
 	eventInsertQuery := m.qf.ConsensusEventInsertQuery()
 
 	for _, event := range data.Events {
@@ -592,7 +596,7 @@ func (m *Main) queueRegistryEvents(batch *storage.QueryBatch, data *storage.Regi
 	return nil
 }
 
-func (m *Main) queueRootHashEvents(batch *storage.QueryBatch, data *storage.RootHashData) error {
+func (m *Main) queueRootHashEventInserts(batch *storage.QueryBatch, data *storage.RootHashData) error {
 	eventInsertQuery := m.qf.ConsensusEventInsertQuery()
 
 	for _, event := range data.Events {
@@ -618,11 +622,47 @@ func (m *Main) queueRootHashEvents(batch *storage.QueryBatch, data *storage.Root
 	return nil
 }
 
-func (m *Main) queueTransfers(batch *storage.QueryBatch, data *storage.StakingData) error {
+// Enum of transfer types. We single out transfers that deduct from the special
+// "fee accumulator" account. These deductions/disbursements happen at the end
+// of each block. However, oasis-core returns each block's events in the
+// following order: BeginBlockEvents, EndBlockEvents (which include
+// disbursements), TxEvents (which fill the fee accumulator). Thus, processing
+// the events in order results in a temporary negative balance for the fee
+// accumulator, which violates our DB checks. We therefore artificially split
+// transfer events into two: accumulator disbursements, and all others. We
+// process the former at the very end.
+// We might be able to remove this once https://github.com/oasisprotocol/oasis-core/pull/5117
+// is deployed, making oasis-core send the "correct" event order on its own. But
+// Cobalt (pre-Damask network) will never be fixed.
+type TransferType string
+
+const (
+	TransferTypeAccumulatorDisbursement TransferType = "AccumulatorDisbursement"
+	TransferTypeOther                   TransferType = "Other"
+)
+
+func (m *Main) queueRegularTransfers(batch *storage.QueryBatch, data *storage.StakingData) error {
+	return m.queueTransfers(batch, data, TransferTypeOther)
+}
+
+func (m *Main) queueDisbursementTransfers(batch *storage.QueryBatch, data *storage.StakingData) error {
+	return m.queueTransfers(batch, data, TransferTypeAccumulatorDisbursement)
+}
+
+func (m *Main) queueTransfers(batch *storage.QueryBatch, data *storage.StakingData, targetType TransferType) error {
 	senderUpdateQuery := m.qf.ConsensusSenderUpdateQuery()
 	receiverUpsertQuery := m.qf.ConsensusReceiverUpdateQuery()
 
 	for _, transfer := range data.Transfers {
+		// Filter out transfers that are not of the target type.
+		typ := TransferTypeOther // type of the current transfer
+		if transfer.From == staking.FeeAccumulatorAddress {
+			typ = TransferTypeAccumulatorDisbursement
+		}
+		if typ != targetType {
+			continue
+		}
+
 		batch.Queue(senderUpdateQuery,
 			transfer.From.String(),
 			transfer.Amount.String(),
@@ -753,7 +793,7 @@ func (m *Main) queueAllowanceChanges(batch *storage.QueryBatch, data *storage.St
 	return nil
 }
 
-func (m *Main) queueStakingEvents(batch *storage.QueryBatch, data *storage.StakingData) error {
+func (m *Main) queueStakingEventInserts(batch *storage.QueryBatch, data *storage.StakingData) error {
 	eventInsertQuery := m.qf.ConsensusEventInsertQuery()
 
 	for _, event := range data.Events {
@@ -894,7 +934,7 @@ func (m *Main) queueVotes(batch *storage.QueryBatch, data *storage.GovernanceDat
 	return nil
 }
 
-func (m *Main) queueGovernanceEvents(batch *storage.QueryBatch, data *storage.GovernanceData) error {
+func (m *Main) queueGovernanceEventInserts(batch *storage.QueryBatch, data *storage.GovernanceData) error {
 	eventInsertQuery := m.qf.ConsensusEventInsertQuery()
 
 	for _, event := range data.Events {