Retry a failed self-check on another validator; add the self-check failure ceiling

A validator that failed the self-check of an address no longer gets that address
again in the current round (ClaimNextQueued skips it); the validator itself stays
in service and takes all other addresses. The verdict fail is set when the number
of failed self-checks of an address reaches settings.self_check_max_attempts
(1..50, default 5, independent of the number of validators); max_retries and
retry_count are no longer used for self-check. If every working validator has
already failed the address, a new round starts and the exclusions lapse.

Migration 0012: ip_self_check_failures (permanent history per registry address),
ip_queue.sc_failures and sc_round_start_cycle (cycle_id is used instead of
attempt_number, which restarts when a queue row is recreated), the setting.
db.FailSelfCheck does it in one transaction; re-submission starts a new series.
API: self_check_max_attempts in GET/PUT /admin/config/orchestrator,
self_check_failed_on in /admin/ips/{ip} and /admin/registry/{ip}. Dashboard: the
field on /settings and the line "Self-check не прошёл на: ..." on the address
pages. Docs, plan and summary in docs/changes/.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

internal/db/db.go

@@ -46,6 +46,9 @@ var verdictIntegritySchema string
 //go:embed migrations/0011_check_runs.sql
 var checkRunsSchema string
 
+//go:embed migrations/0012_self_check_failures.sql
+var selfCheckFailuresSchema string
+
 // migrations is the ordered list of schema versions. Each entry's SQL is
 // applied, in order, for any version greater than the database's current
 // PRAGMA user_version — so a fresh database walks the whole list and an
@@ -65,6 +68,7 @@ var migrations = []struct {
 	{9, scaleIndexesSchema},
 	{10, verdictIntegritySchema},
 	{11, checkRunsSchema},
+	{12, selfCheckFailuresSchema},
 }
 
 type DB struct {

internal/db/migrations/0012_self_check_failures.sql

@@ -0,0 +1,38 @@
+-- Self-check failures per address (see
+-- docs/changes/2026-10-04_08-01_self-check-exclude-validator-plan.md).
+--
+-- A failed self-check no longer sends the address back to the validator that
+-- failed it: ClaimNextQueued skips an address for every validator that failed
+-- it in the current round. ip_self_check_failures is the permanent history of
+-- those failures; it is keyed by registry_id (like checks and events), so it
+-- outlives the ip_queue row and is not touched by re-checks. No foreign keys
+-- on purpose: the manual cleanup in docs/ADMIN_CLEANUP.md deletes freely.
+--
+-- A round is the part of a series of failures in which validators that failed
+-- stay excluded. ip_queue.sc_round_start_cycle is the first cycle_id of the
+-- current round: a failure excludes its validator only if its cycle_id is not
+-- below it. cycle_id (not attempt_number) is used because it never repeats for
+-- an address, even when the ip_queue row is deleted and created again.
+--
+-- ip_queue.sc_failures counts self-check failures of the current series; a
+-- manual re-check or re-submission starts a new series.
+--
+-- settings.self_check_max_attempts is the ceiling of self-check failures per
+-- address, after which the address gets the verdict fail (1..50, default 5).
+
+CREATE TABLE ip_self_check_failures (
+  id             INTEGER PRIMARY KEY AUTOINCREMENT,
+  registry_id    INTEGER NOT NULL,
+  run_id         INTEGER,
+  cycle_id       INTEGER NOT NULL,
+  attempt_number INTEGER NOT NULL,
+  validator_id   TEXT NOT NULL,
+  failed_at      TIMESTAMP NOT NULL,
+  detail         TEXT NOT NULL DEFAULT ''
+);
+CREATE INDEX idx_sc_failures_registry ON ip_self_check_failures(registry_id, validator_id);
+
+ALTER TABLE ip_queue ADD COLUMN sc_failures INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE ip_queue ADD COLUMN sc_round_start_cycle INTEGER NOT NULL DEFAULT 0;
+
+ALTER TABLE settings ADD COLUMN self_check_max_attempts INTEGER NOT NULL DEFAULT 5;

internal/db/models.go

@@ -276,10 +276,20 @@ type Settings struct {
 	// HistoryRetentionCycles caps how many recent check cycles are kept per
 	// registry address (see PruneRegistryHistory); 0 means unlimited.
 	HistoryRetentionCycles int
-	CreatedAt              time.Time
-	UpdatedAt              time.Time
+	// SelfCheckMaxAttempts is the ceiling of failed self-checks per address
+	// (1..MaxSelfCheckMaxAttempts); reaching it gives the address the verdict
+	// fail (see FailSelfCheck).
+	SelfCheckMaxAttempts int
+	CreatedAt            time.Time
+	UpdatedAt            time.Time
 }
 
+// Bounds of Settings.SelfCheckMaxAttempts.
+const (
+	MinSelfCheckMaxAttempts = 1
+	MaxSelfCheckMaxAttempts = 50
+)
+
 // InboundChecksSettings is the singleton row describing what the prober
 // checks on every site for every in-flight IP (TCP ports + optional ICMP).
 // Admin-configurable at runtime (see queries_inbound.go).

internal/db/queries_ipqueue.go

@@ -44,10 +44,10 @@ func (d *DB) SeedQueue(ctx context.Context, addresses []string) error {
 			}
 		}
 		if _, err := tx.ExecContext(ctx, `
-			INSERT INTO ip_queue (ip_address, sequence, state, registry_id, cycle_id, run_id, created_at, updated_at)
-			VALUES (?, ?, ?, ?, ?, ?, ?, ?)
+			INSERT INTO ip_queue (ip_address, sequence, state, registry_id, cycle_id, run_id, sc_round_start_cycle, created_at, updated_at)
+			VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
 			ON CONFLICT(ip_address) DO NOTHING
-		`, addr, i, IPQueued, registryID, cycle, runID, now, now); err != nil {
+		`, addr, i, IPQueued, registryID, cycle, runID, cycle, now, now); err != nil {
 			return fmt.Errorf("seed %s: %w", addr, err)
 		}
 	}
@@ -55,8 +55,11 @@ func (d *DB) SeedQueue(ctx context.Context, addresses []string) error {
 }
 
 // ClaimNextQueued atomically hands the next queued IP (lowest sequence) to
-// the given idle validator. It returns (nil, nil) if the validator isn't
-// idle or no IP is queued. The DB connection pool is capped at one physical
+// the given idle validator, skipping addresses this validator failed the
+// self-check of in the current round (see FailSelfCheck): such an address
+// stays queued for the other validators and does not block the ones behind
+// it. It returns (nil, nil) if the validator isn't idle or no IP is
+// claimable for it. The DB connection pool is capped at one physical
 // connection (see Open), so this transaction already has exclusive access
 // to the database for its duration — no other claim, requeue, or update can
 // interleave — which combined with the conditional UPDATEs (checked via
@@ -82,9 +85,12 @@ func (d *DB) ClaimNextQueued(ctx context.Context, validatorID string, leaseTTL t
 
 	var item IPQueueItem
 	err = tx.QueryRowContext(ctx, `
-		SELECT id, ip_address, sequence, attempt_number, retry_count
-		FROM ip_queue WHERE state=? ORDER BY sequence LIMIT 1
-	`, IPQueued).Scan(&item.ID, &item.IPAddress, &item.Sequence, &item.AttemptNumber, &item.RetryCount)
+		SELECT q.id, q.ip_address, q.sequence, q.attempt_number, q.retry_count
+		FROM ip_queue q WHERE q.state=? AND NOT EXISTS (
+			SELECT 1 FROM ip_self_check_failures f
+			WHERE f.registry_id=q.registry_id AND f.validator_id=? AND f.cycle_id>=q.sc_round_start_cycle)
+		ORDER BY q.sequence LIMIT 1
+	`, IPQueued, validatorID).Scan(&item.ID, &item.IPAddress, &item.Sequence, &item.AttemptNumber, &item.RetryCount)
 	if err == sql.ErrNoRows {
 		return nil, nil
 	}
@@ -352,6 +358,178 @@ func (d *DB) RequeueOrFail(ctx context.Context, ipID int64, validatorID string,
 	return tx.Commit()
 }
 
+// SelfCheckFailure is the outcome of FailSelfCheck.
+type SelfCheckFailure struct {
+	// Failures is the number of failed self-checks in the address's current
+	// series, this one included.
+	Failures int
+	// Failed: the ceiling was reached and the address got the verdict fail.
+	Failed bool
+	// Validators lists the validators that failed the self-check in this
+	// series, oldest first (with repeats if one failed it more than once).
+	Validators []string
+	// NewRound: the address was queued again, and every working validator had
+	// already failed it in the round, so the round was reset — the exclusions
+	// no longer apply (always false when Failed).
+	NewRound bool
+}
+
+// FailSelfCheck handles a failed self-check of the address ipID held by
+// validatorID, in one transaction: it records the failure in
+// ip_self_check_failures and in ip_queue.sc_failures, then
+//
+//   - if sc_failures reached maxAttempts: marks the address failed with the
+//     verdict fail;
+//   - otherwise sends it back to the queue without touching retry_count (a
+//     failed self-check has its own ceiling, unlike a failed association or
+//     an expired lease, see RequeueOrFail). The validator is excluded from
+//     this address for the rest of the round (ClaimNextQueued). If no
+//     working validator (idle, assigned, checking) is left without a failure
+//     in the round, a new round starts: the exclusions lapse and the retry
+//     follows the usual rules, so with fewer validators than maxAttempts the
+//     address never gets stuck in the queue.
+//
+// The validator is freed in the same transaction. Returns ErrInvalidState if
+// the address is not awaiting_self_check on this validator (a late report).
+func (d *DB) FailSelfCheck(ctx context.Context, ipID int64, validatorID, detail string, maxAttempts int) (SelfCheckFailure, error) {
+	var out SelfCheckFailure
+	tx, err := d.BeginTx(ctx, nil)
+	if err != nil {
+		return out, err
+	}
+	defer tx.Rollback()
+
+	var registryID, cycle, roundStart int64
+	var runID sql.NullInt64
+	var attempt, failures int
+	err = tx.QueryRowContext(ctx, `
+		SELECT registry_id, run_id, cycle_id, attempt_number, sc_failures, sc_round_start_cycle
+		FROM ip_queue WHERE id=? AND state=? AND owner_validator_id=?
+	`, ipID, IPAwaitingSelfCheck, validatorID).Scan(&registryID, &runID, &cycle, &attempt, &failures, &roundStart)
+	if err == sql.ErrNoRows {
+		return out, fmt.Errorf("ip_id %d is not awaiting self-check on %s: %w", ipID, validatorID, ErrInvalidState)
+	}
+	if err != nil {
+		return out, err
+	}
+
+	now := timeToDB(Now())
+	if _, err := tx.ExecContext(ctx, `
+		INSERT INTO ip_self_check_failures (registry_id, run_id, cycle_id, attempt_number, validator_id, failed_at, detail)
+		VALUES (?, ?, ?, ?, ?, ?, ?)
+	`, registryID, runID, cycle, attempt, validatorID, now, detail); err != nil {
+		return out, fmt.Errorf("record self-check failure: %w", err)
+	}
+	failures++
+	out.Failures = failures
+
+	rows, err := tx.QueryContext(ctx, `
+		SELECT validator_id FROM (
+			SELECT id, validator_id FROM ip_self_check_failures WHERE registry_id=? ORDER BY id DESC LIMIT ?
+		) ORDER BY id
+	`, registryID, failures)
+	if err != nil {
+		return out, err
+	}
+	for rows.Next() {
+		var v string
+		if err := rows.Scan(&v); err != nil {
+			rows.Close()
+			return out, err
+		}
+		out.Validators = append(out.Validators, v)
+	}
+	if err := rows.Err(); err != nil {
+		rows.Close()
+		return out, err
+	}
+	rows.Close()
+
+	if failures >= maxAttempts {
+		out.Failed = true
+		if _, err := tx.ExecContext(ctx, `
+			UPDATE ip_queue SET
+				state=?, sc_failures=?, overall_result=?, aggregated_at=?, updated_at=?
+			WHERE id=?
+		`, IPFailed, failures, ResultFail, now, now, ipID); err != nil {
+			return out, err
+		}
+		if err := upsertRunResultTx(ctx, tx, ipID, ResultFail, -1, now); err != nil {
+			return out, err
+		}
+		if err := finalizeRunsTx(ctx, tx, now); err != nil {
+			return out, err
+		}
+	} else {
+		newCycle, err := nextRegistryCycleTx(ctx, tx, registryID, now)
+		if err != nil {
+			return out, err
+		}
+		var left int
+		if err := tx.QueryRowContext(ctx, `
+			SELECT COUNT(*) FROM validators v
+			WHERE v.state IN (?, ?, ?) AND NOT EXISTS (
+				SELECT 1 FROM ip_self_check_failures f
+				WHERE f.registry_id=? AND f.validator_id=v.validator_id AND f.cycle_id>=?)
+		`, ValidatorIdle, ValidatorAssigned, ValidatorChecking, registryID, roundStart).Scan(&left); err != nil {
+			return out, err
+		}
+		if left == 0 {
+			out.NewRound = true
+			roundStart = int64(newCycle)
+		}
+		if _, err := tx.ExecContext(ctx, `
+			UPDATE ip_queue SET
+				state=?, owner_validator_id=NULL, fip_id='', attempt_number=attempt_number+1,
+				cycle_id=?, sc_failures=?, sc_round_start_cycle=?, lease_expires_at=NULL, egress_complete=0,
+				overall_result='', assigned_at=NULL, checking_started_at=NULL, fip_associated_at=NULL, updated_at=?
+			WHERE id=?
+		`, IPQueued, newCycle, failures, roundStart, now, ipID); err != nil {
+			return out, err
+		}
+	}
+
+	if _, err := tx.ExecContext(ctx, freeValidatorSQL, now, validatorID, ipID); err != nil {
+		return out, err
+	}
+	return out, tx.Commit()
+}
+
+// GetIPRunID returns the check run the queue row belongs to (0 if none).
+func (d *DB) GetIPRunID(ctx context.Context, ipID int64) (int64, error) {
+	var runID sql.NullInt64
+	if err := d.QueryRowContext(ctx, `SELECT run_id FROM ip_queue WHERE id=?`, ipID).Scan(&runID); err != nil {
+		return 0, err
+	}
+	return runID.Int64, nil
+}
+
+// ListSelfCheckFailedOn returns the distinct validators whose self-check of
+// the address failed, in alphabetical order: over the whole history of the
+// address, or, if runID > 0, only the failures that happened in that run.
+func (d *DB) ListSelfCheckFailedOn(ctx context.Context, registryID, runID int64) ([]string, error) {
+	q := `SELECT DISTINCT validator_id FROM ip_self_check_failures WHERE registry_id=?`
+	args := []any{registryID}
+	if runID > 0 {
+		q += ` AND run_id=?`
+		args = append(args, runID)
+	}
+	rows, err := d.QueryContext(ctx, q+` ORDER BY validator_id`, args...)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	out := []string{}
+	for rows.Next() {
+		var v string
+		if err := rows.Scan(&v); err != nil {
+			return nil, err
+		}
+		out = append(out, v)
+	}
+	return out, rows.Err()
+}
+
 // SubmitIPs is the single admin entry point for both "add new addresses to
 // the queue" and "force a re-check of an already-finished address" — the
 // same list can freely mix both. Addresses are processed in one
@@ -360,7 +538,9 @@ func (d *DB) RequeueOrFail(ctx context.Context, ipID int64, validatorID string,
 //   - unknown address: inserted as a new queued row.
 //   - address currently done/failed/occupied: reset to queued (new attempt,
 //     retry_count cleared — this is a deliberate admin-triggered restart,
-//     not a system retry).
+//     not a system retry). It also starts a new series of self-check
+//     failures (sc_failures=0, validators that failed it before are no
+//     longer excluded); the history in ip_self_check_failures stays.
 //   - address currently queued (not yet claimed): left in state=queued,
 //     only its sequence is updated.
 //   - address currently mid-check (assigning_fip / awaiting_self_check /
@@ -426,9 +606,9 @@ func (d *DB) SubmitIPsAs(ctx context.Context, addresses []string, kind string) (
 				return result, rErr
 			}
 			if _, err := tx.ExecContext(ctx, `
-				INSERT INTO ip_queue (ip_address, sequence, state, registry_id, cycle_id, run_id, created_at, updated_at)
-				VALUES (?, ?, ?, ?, ?, ?, ?, ?)
-			`, addr, seq, IPQueued, registryID, cycle, rid, now, now); err != nil {
+				INSERT INTO ip_queue (ip_address, sequence, state, registry_id, cycle_id, run_id, sc_round_start_cycle, created_at, updated_at)
+				VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+			`, addr, seq, IPQueued, registryID, cycle, rid, cycle, now, now); err != nil {
 				return result, fmt.Errorf("insert %s: %w", addr, err)
 			}
 			result.Added = append(result.Added, addr)
@@ -453,10 +633,10 @@ func (d *DB) SubmitIPsAs(ctx context.Context, addresses []string, kind string) (
 				UPDATE ip_queue SET
 					state=?, sequence=?, owner_validator_id=NULL, fip_id='', retry_count=0,
 					attempt_number=attempt_number+1, cycle_id=?, lease_expires_at=NULL, egress_complete=0,
-					overall_result='', run_id=?,
+					overall_result='', run_id=?, sc_failures=0, sc_round_start_cycle=?,
 					assigned_at=NULL, checking_started_at=NULL, fip_associated_at=NULL, aggregated_at=NULL, fip_released_at=NULL, updated_at=?
 				WHERE ip_address=?
-			`, IPQueued, seq, cycle, rid, now, addr); err != nil {
+			`, IPQueued, seq, cycle, rid, cycle, now, addr); err != nil {
 				return result, fmt.Errorf("requeue %s: %w", addr, err)
 			}
 			result.Requeued = append(result.Requeued, addr)

internal/db/queries_runs_test.go

@@ -383,7 +383,7 @@ func TestMigration0011BuildsRunsFromExistingData(t *testing.T) {
 	}
 	var ver int
 	d.QueryRowContext(ctx, `PRAGMA user_version`).Scan(&ver)
-	if ver != 11 {
+	if ver != 12 {
 		t.Errorf("user_version = %d", ver)
 	}
 }

internal/db/queries_selfcheck_test.go

@@ -0,0 +1,190 @@
+package db
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"cloudipvalidator/internal/config"
+)
+
+// claimAwaitingSelfCheck claims the next address for the validator and brings
+// it to awaiting_self_check, as the orchestrator does after the association.
+func claimAwaitingSelfCheck(t *testing.T, ctx context.Context, d *DB, validatorID string) *IPQueueItem {
+	t.Helper()
+	item, err := d.ClaimNextQueued(ctx, validatorID, time.Minute)
+	if err != nil || item == nil {
+		t.Fatalf("claim for %s: item=%+v err=%v", validatorID, item, err)
+	}
+	if err := d.SetFIPAssociated(ctx, item.ID, "fip-"+item.IPAddress, time.Minute); err != nil {
+		t.Fatalf("set fip associated: %v", err)
+	}
+	return item
+}
+
+// A validator that failed the self-check of an address is not given that
+// address again, takes the next one instead, and another validator takes the
+// excluded one; the exclusion covers that address only.
+func TestClaimSkipsAddressExcludedForValidator(t *testing.T) {
+	d, ctx := newTestDB(t)
+	if err := d.SeedQueue(ctx, []string{"1.1.1.1", "2.2.2.2"}); err != nil {
+		t.Fatal(err)
+	}
+	for _, v := range []string{"v1", "v2"} {
+		if err := d.AdminCreateValidator(ctx, v, "port-"+v); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	first := claimAwaitingSelfCheck(t, ctx, d, "v1")
+	if first.IPAddress != "1.1.1.1" {
+		t.Fatalf("expected 1.1.1.1 first, got %s", first.IPAddress)
+	}
+	res, err := d.FailSelfCheck(ctx, first.ID, "v1", "ip echo timeout", 5)
+	if err != nil {
+		t.Fatalf("fail self-check: %v", err)
+	}
+	if res.Failed || res.NewRound || res.Failures != 1 {
+		t.Fatalf("expected a plain retry after the first failure, got %+v", res)
+	}
+	back, _ := d.GetIP(ctx, first.ID)
+	if back.State != IPQueued || back.RetryCount != 0 {
+		t.Fatalf("expected queued with retry_count untouched, got state=%s retry_count=%d", back.State, back.RetryCount)
+	}
+	if v, _ := d.GetValidator(ctx, "v1"); v.State != ValidatorIdle {
+		t.Fatalf("expected v1 freed to idle, got %s", v.State)
+	}
+
+	// v1 skips 1.1.1.1 (still ahead in the queue) and takes 2.2.2.2.
+	second, err := d.ClaimNextQueued(ctx, "v1", time.Minute)
+	if err != nil || second == nil || second.IPAddress != "2.2.2.2" {
+		t.Fatalf("expected v1 to take 2.2.2.2, got %+v err=%v", second, err)
+	}
+	// v2 takes the excluded address.
+	other, err := d.ClaimNextQueued(ctx, "v2", time.Minute)
+	if err != nil || other == nil || other.IPAddress != "1.1.1.1" {
+		t.Fatalf("expected v2 to take 1.1.1.1, got %+v err=%v", other, err)
+	}
+}
+
+// When every working validator has failed the address in the round, a new
+// round starts and the exclusions lapse; an unreachable validator does not
+// count as working.
+func TestSelfCheckNewRoundLiftsExclusions(t *testing.T) {
+	d, ctx := newTestDB(t)
+	if err := d.SeedQueue(ctx, []string{"1.1.1.1"}); err != nil {
+		t.Fatal(err)
+	}
+	for _, v := range []string{"v1", "v2"} {
+		if err := d.AdminCreateValidator(ctx, v, "port-"+v); err != nil {
+			t.Fatal(err)
+		}
+	}
+	if _, err := d.Exec(`UPDATE validators SET state=? WHERE validator_id='v2'`, ValidatorUnreachable); err != nil {
+		t.Fatal(err)
+	}
+
+	item := claimAwaitingSelfCheck(t, ctx, d, "v1")
+	res, err := d.FailSelfCheck(ctx, item.ID, "v1", "x", 5)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !res.NewRound {
+		t.Fatalf("expected a new round (the only working validator failed it), got %+v", res)
+	}
+	again, err := d.ClaimNextQueued(ctx, "v1", time.Minute)
+	if err != nil || again == nil || again.IPAddress != "1.1.1.1" {
+		t.Fatalf("expected v1 to get the address again in the new round, got %+v err=%v", again, err)
+	}
+}
+
+// Reaching the ceiling gives the verdict fail; a re-submission starts a new
+// series (counter and exclusions) but keeps the failure history.
+func TestSelfCheckCeilingAndResubmit(t *testing.T) {
+	d, ctx := newTestDB(t)
+	if err := d.SeedQueue(ctx, []string{"1.1.1.1"}); err != nil {
+		t.Fatal(err)
+	}
+	for _, v := range []string{"v1", "v2"} {
+		if err := d.AdminCreateValidator(ctx, v, "port-"+v); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	item := claimAwaitingSelfCheck(t, ctx, d, "v1")
+	if res, err := d.FailSelfCheck(ctx, item.ID, "v1", "x", 2); err != nil || res.Failed {
+		t.Fatalf("below the ceiling: res=%+v err=%v", res, err)
+	}
+	item = claimAwaitingSelfCheck(t, ctx, d, "v2")
+	res, err := d.FailSelfCheck(ctx, item.ID, "v2", "y", 2)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !res.Failed || res.Failures != 2 || len(res.Validators) != 2 || res.Validators[0] != "v1" || res.Validators[1] != "v2" {
+		t.Fatalf("expected fail at the ceiling on v1, v2, got %+v", res)
+	}
+	ip, _ := d.GetIP(ctx, item.ID)
+	if ip.State != IPFailed || ip.OverallResult != ResultFail {
+		t.Fatalf("expected failed/fail, got %s/%s", ip.State, ip.OverallResult)
+	}
+
+	if _, err := d.SubmitIPs(ctx, []string{"1.1.1.1"}); err != nil {
+		t.Fatal(err)
+	}
+	var failures int
+	if err := d.QueryRow(`SELECT sc_failures FROM ip_queue WHERE id=?`, item.ID).Scan(&failures); err != nil || failures != 0 {
+		t.Fatalf("expected the series reset to 0, got %d err=%v", failures, err)
+	}
+	if got, err := d.ListSelfCheckFailedOn(ctx, ip.RegistryID, 0); err != nil || len(got) != 2 || got[0] != "v1" || got[1] != "v2" {
+		t.Fatalf("expected the history kept (v1, v2), got %v err=%v", got, err)
+	}
+	// v1 failed it before the re-submission, but is no longer excluded.
+	if again, err := d.ClaimNextQueued(ctx, "v1", time.Minute); err != nil || again == nil {
+		t.Fatalf("expected v1 to claim the re-submitted address, got %+v err=%v", again, err)
+	}
+}
+
+// A late report for an address the validator does not hold is refused.
+func TestFailSelfCheckRequiresAwaitingSelfCheckOnValidator(t *testing.T) {
+	d, ctx := newTestDB(t)
+	if err := d.SeedQueue(ctx, []string{"1.1.1.1"}); err != nil {
+		t.Fatal(err)
+	}
+	for _, v := range []string{"v1", "v2"} {
+		if err := d.AdminCreateValidator(ctx, v, "port-"+v); err != nil {
+			t.Fatal(err)
+		}
+	}
+	item := claimAwaitingSelfCheck(t, ctx, d, "v1")
+	if _, err := d.FailSelfCheck(ctx, item.ID, "v2", "late", 5); !errors.Is(err, ErrInvalidState) {
+		t.Fatalf("expected ErrInvalidState for another validator, got %v", err)
+	}
+	cur, _ := d.GetIP(ctx, item.ID)
+	if got, _ := d.ListSelfCheckFailedOn(ctx, cur.RegistryID, 0); len(got) != 0 {
+		t.Fatalf("a refused report must leave no history, got %v", got)
+	}
+}
+
+func TestSelfCheckMaxAttemptsSetting(t *testing.T) {
+	d, ctx := newTestDB(t)
+	if err := d.BootstrapFromConfig(ctx, &config.ControlAPI{}); err != nil {
+		t.Fatal(err)
+	}
+	if s, err := d.GetSettings(ctx); err != nil || s.SelfCheckMaxAttempts != 5 {
+		t.Fatalf("expected default 5, got %+v err=%v", s, err)
+	}
+	for _, bad := range []int{0, -1, 51} {
+		if err := d.SetSelfCheckMaxAttempts(ctx, bad); !errors.Is(err, ErrValidation) {
+			t.Fatalf("expected ErrValidation for %d, got %v", bad, err)
+		}
+	}
+	for _, ok := range []int{1, 50} {
+		if err := d.SetSelfCheckMaxAttempts(ctx, ok); err != nil {
+			t.Fatalf("set %d: %v", ok, err)
+		}
+		if s, _ := d.GetSettings(ctx); s.SelfCheckMaxAttempts != ok {
+			t.Fatalf("expected %d, got %d", ok, s.SelfCheckMaxAttempts)
+		}
+	}
+}

internal/db/queries_settings.go

@@ -13,8 +13,8 @@ func (d *DB) GetSettings(ctx context.Context) (Settings, error) {
 	var s Settings
 	var createdAt, updatedAt string
 	err := d.QueryRowContext(ctx, `
-		SELECT fip_settle_seconds, history_retention_cycles, created_at, updated_at FROM settings WHERE id=1
-	`).Scan(&s.FIPSettleSeconds, &s.HistoryRetentionCycles, &createdAt, &updatedAt)
+		SELECT fip_settle_seconds, history_retention_cycles, self_check_max_attempts, created_at, updated_at FROM settings WHERE id=1
+	`).Scan(&s.FIPSettleSeconds, &s.HistoryRetentionCycles, &s.SelfCheckMaxAttempts, &createdAt, &updatedAt)
 	if err != nil {
 		return Settings{}, err
 	}
@@ -58,3 +58,17 @@ func (d *DB) SetHistoryRetentionCycles(ctx context.Context, cycles int) error {
 	`, cycles, now)
 	return err
 }
+
+// SetSelfCheckMaxAttempts persists the ceiling of failed self-checks per
+// address (1..MaxSelfCheckMaxAttempts). It applies from the next failed
+// self-check, without a restart.
+func (d *DB) SetSelfCheckMaxAttempts(ctx context.Context, attempts int) error {
+	if attempts < MinSelfCheckMaxAttempts || attempts > MaxSelfCheckMaxAttempts {
+		return fmt.Errorf("self_check_max_attempts must be in %d..%d: %w", MinSelfCheckMaxAttempts, MaxSelfCheckMaxAttempts, ErrValidation)
+	}
+	now := timeToDB(Now())
+	_, err := d.ExecContext(ctx, `
+		UPDATE settings SET self_check_max_attempts=?, updated_at=? WHERE id=1
+	`, attempts, now)
+	return err
+}

internal/db/queries_verdict_integrity_test.go

@@ -240,7 +240,7 @@ func TestMigration0010MarksRowsAfterVerdict(t *testing.T) {
 		t.Errorf("ssh: after_verdict=%d recorded=%s created=%s", a, rec, cr)
 	}
 	var ver int
-	if err := d.QueryRowContext(ctx, `PRAGMA user_version`).Scan(&ver); err != nil || ver != 11 {
+	if err := d.QueryRowContext(ctx, `PRAGMA user_version`).Scan(&ver); err != nil || ver != 12 {
 		t.Errorf("user_version=%d err=%v", ver, err)
 	}
 }