diff --git a/README.md b/README.md
index 82c4b53..a34dfa2 100644
--- a/README.md
+++ b/README.md
@@ -8,7 +8,7 @@ Fake pod implementation
 
 * This version also mimics a behavior we see in some DASH pods where the pod disconnects every 3 minutes; this can be used with iOS hooks to make a heartbeat to run Loop in situations where a BLE CGM is not available. 
 
-* It serves a control panel at http://<host>:8080/ (`-api` changes the address): the pod's reservoir, delivery, age, alerts and fault, with controls to set the reservoir and pod age, raise alerts, trigger or clear a fault, and crash on the next command to test unacknowledged commands. The panel drives a JSON API: `GET /status`, and `POST /command` with `{"command": "changeReservoir", "value": 45}` (also `setActiveTime` in minutes, `setAlerts` as a slot mask, `setFault` with a fault code, 0 to clear, `crashNextCommand` with `"beforeProcessing": true|false`, `ignoreNextMessages` with a count, which keeps the link up but leaves that many messages unanswered, as a pod that didn't hear them would, and `failNextConnects` with a count and an optional `hangUpAfter` in seconds, 7 by default, which leaves service discovery unanswered on that many connects and then hangs up with CoreBluetooth's "peripheral disconnected" error, as some pods do in the field; this needs `-transport impossible` and an ImpossiBLE that passes `errorDomain` and `errorCode` through on a disconnect). A fault stops a bolus in progress, and the pod reports the part left undelivered however late it is read, as a real pod does.
+* It serves a control panel at http://<host>:8080/ (`-api` changes the address): the pod's reservoir, delivery, age, alerts and fault, with controls to set the reservoir and pod age, raise alerts, trigger or clear a fault, and crash on the next command to test unacknowledged commands. The panel drives a JSON API: `GET /status`, and `POST /command` with `{"command": "changeReservoir", "value": 45}` (also `setActiveTime` in minutes, `setAlerts` as a slot mask, `setFault` with a fault code, 0 to clear, `crashNextCommand` with `"beforeProcessing": true|false`, `ignoreNextMessages` with a count, which keeps the link up but leaves that many messages unanswered, as a pod that didn't hear them would, and `failNextConnects` with a count and an optional `hangUpAfter` in seconds, 7 by default, which leaves service discovery unanswered on that many connects and then hangs up with CoreBluetooth's "peripheral disconnected" error, as some pods do in the field; this needs `-transport impossible` and an ImpossiBLE that passes `errorDomain` and `errorCode` through on a disconnect). A fault stops a bolus in progress, and the pod reports the part left undelivered however late it is read, as a real pod does. A reset fault (0x0D, 0x0E, 0x0F, 0x10, 0x12, 0x22, 0x34, 0x3B, 0xAA, 0xD6 or 0xD7) reports what a reset pod does instead: no fault time, nothing delivered or left undelivered, and minutes active counted from the reset.
 
 * The websocket API at `/ws`, used by the separate [NodeJS/React frontend](https://github.com/ps2/pod_simulator_frontend), is still there.
 
diff --git a/pkg/pod/pod.go b/pkg/pod/pod.go
index 42e9af5..55fe4f9 100644
--- a/pkg/pod/pod.go
+++ b/pkg/pod/pod.go
@@ -598,6 +598,7 @@ func (p *Pod) makeGeneralStatusResponse() response.Response {
 func (p *Pod) makeDetailedStatusResponse() response.Response {
 
 	var now = time.Now()
+	delivered, bolusRemaining := p.state.ReportedDelivery()
 
 	return &response.DetailedStatusResponse{
 		LastProgSeqNum:      p.state.LastProgSeqNum,
@@ -608,9 +609,9 @@ func (p *Pod) makeDetailedStatusResponse() response.Response {
 		BasalActive:         p.state.BasalActive,
 		ExtendedBolusActive: p.state.ExtendedBolusActive,
 		PodProgress:         p.state.PodProgress,
-		Delivered:           p.state.Delivered,
-		BolusRemaining:      p.state.BolusRemaining(),
-		MinutesActive:       p.state.MinutesActive(),
+		Delivered:           delivered,
+		BolusRemaining:      bolusRemaining,
+		MinutesActive:       p.state.ReportedMinutesActive(),
 		FaultEvent:          p.state.FaultEvent,
 		FaultEventTime:      p.state.FaultTime,
 	}
@@ -628,7 +629,7 @@ func (p *Pod) makeType3StatusResponse() response.Response {
 	return &response.Type3StatusResponse{
 		FaultEvent:     p.state.FaultEvent,
 		FaultEventTime: p.state.FaultTime,
-		MinutesActive:  p.state.MinutesActive(),
+		MinutesActive:  p.state.ReportedMinutesActive(),
 	}
 }
 
diff --git a/pkg/pod/state.go b/pkg/pod/state.go
index 4b7756d..7f4bd9b 100644
--- a/pkg/pod/state.go
+++ b/pkg/pod/state.go
@@ -39,6 +39,9 @@ type PODState struct {
 	// pod keeps reporting it, however late it is read.
 	FaultBolusRemaining uint16 `toml:"fault_bolus_remaining"`
 
+	// ResetAt is when a reset fault restarted the pod's counters.
+	ResetAt time.Time `toml:"reset_at"`
+
 	TriggerTimes [8]uint16 `toml:"trigger_times"`
 
 	// At some point these could be replaced with details
@@ -134,6 +137,42 @@ func (p *PODState) SetFault(code uint8) {
 	}
 	p.FaultEvent = code
 	p.FaultTime = p.MinutesActive()
+	if IsResetFault(code) {
+		if p.ResetAt.IsZero() {
+			p.ResetAt = time.Now()
+		}
+		p.FaultTime = 0
+	} else {
+		p.ResetAt = time.Time{}
+	}
+}
+
+// IsResetFault is a fault that resets the pod. Its status then holds no
+// fault time, nothing delivered or left undelivered, and minutes counted
+// from the reset, as every reset in field reports does.
+func IsResetFault(code uint8) bool {
+	switch code {
+	case 0x0D, 0x0E, 0x0F, 0x10, 0x12, 0x22, 0x34, 0x3B, 0xAA, 0xD6, 0xD7:
+		return true
+	}
+	return false
+}
+
+// ReportedMinutesActive is the pod's minute counter, which a reset restarts.
+func (p *PODState) ReportedMinutesActive() uint16 {
+	if !p.ResetAt.IsZero() {
+		return uint16(time.Now().Sub(p.ResetAt).Round(time.Minute).Minutes())
+	}
+	return p.MinutesActive()
+}
+
+// ReportedDelivery is the delivered total and bolus remaining the pod
+// reports, both of which a reset clears.
+func (p *PODState) ReportedDelivery() (delivered uint16, bolusRemaining uint16) {
+	if !p.ResetAt.IsZero() {
+		return 0, 0
+	}
+	return p.Delivered, p.BolusRemaining()
 }
 
 // NOTE: only handles immediate boluses; any extended bolus is not accounted for
diff --git a/pkg/pod/state_test.go b/pkg/pod/state_test.go
index ca88874..a77726c 100644
--- a/pkg/pod/state_test.go
+++ b/pkg/pod/state_test.go
@@ -1,6 +1,7 @@
 package pod
 
 import (
+	"encoding/hex"
 	"testing"
 	"time"
 
@@ -55,3 +56,55 @@ func TestAFaultWithNoBolusLeavesNothingUndelivered(t *testing.T) {
 		t.Fatalf("remaining %d, delivered %d; want 0 and 1080", got, s.Delivered)
 	}
 }
+
+func TestAResetFaultClearsWhatThePodReports(t *testing.T) {
+	// 3 U (60 pulses) bolus, 20 s in, on a pod 18 minutes old
+	s := &PODState{
+		PodProgress:    response.PodProgressRunningAbove50U,
+		ActivationTime: time.Now().Add(-18 * time.Minute),
+		Delivered:      149,
+		Reservoir:      2851,
+		BolusEnd:       time.Now().Add(100 * time.Second),
+	}
+	s.SetFault(0x10)
+
+	p := &Pod{state: s}
+	raw, _ := p.makeDetailedStatusResponse().Marshal()
+	got := hex.EncodeToString(raw)
+	// no bolus remaining, nothing delivered, fault 0x10 at no time, 0 minutes active
+	want := "0216020d00000000000010000003ff0000"
+	if got[:len(want)] != want {
+		t.Fatalf("reset fault status %s, want it to start %s", got, want)
+	}
+
+	// The pod's own books still have the bolus stopped at the fault
+	if s.Delivered != 99 || s.FaultBolusRemaining != 50 {
+		t.Fatalf("delivered %d, fault bolus remaining %d; want 99 and 50", s.Delivered, s.FaultBolusRemaining)
+	}
+
+	s.ResetAt = time.Now().Add(-3 * time.Minute)
+	if got := s.ReportedMinutesActive(); got != 3 {
+		t.Fatalf("minutes active 3 minutes after the reset = %d, want 3", got)
+	}
+
+	s.SetFault(0)
+	if !s.ResetAt.IsZero() || s.ReportedMinutesActive() != 18 {
+		t.Fatalf("clearing the fault left reset at %v, minutes active %d", s.ResetAt, s.ReportedMinutesActive())
+	}
+}
+
+func TestANonResetFaultKeepsItsTimeAndCounters(t *testing.T) {
+	s := &PODState{
+		PodProgress:    response.PodProgressRunningAbove50U,
+		ActivationTime: time.Now().Add(-100 * time.Minute),
+		Delivered:      1080,
+	}
+	s.SetFault(0x18)
+
+	if s.FaultTime != 100 || !s.ResetAt.IsZero() {
+		t.Fatalf("fault time %d, reset at %v; want 100 and none", s.FaultTime, s.ResetAt)
+	}
+	if delivered, _ := s.ReportedDelivery(); delivered != 1080 || s.ReportedMinutesActive() != 100 {
+		t.Fatalf("reported delivered %d, minutes %d; want 1080 and 100", delivered, s.ReportedMinutesActive())
+	}
+}