fix UI 503s under checkpoint load

- Store: skip identical CT checkpoint writes (CertSpotter calls StoreLogState
  per batch across 64 logs; hashing the state avoids redundant writes).
- webui: cache /api/stats for 10s and raise API query timeouts; UI polls, so
  repeated aggregate queries no longer queue behind checkpoint bursts.
- reword discovery detail (no 'name enumeration' for S3/GCS; only Azure uses a
  container list).

internal/store/state.go

@@ -2,6 +2,7 @@ package store
 
 import (
 	"context"
+	"crypto/sha256"
 	"encoding/json"
 	"fmt"
 	"log"
@@ -35,11 +36,21 @@ func (s *Store) StoreLogState(ctx context.Context, logID monitor.LogID, state *m
 	if err != nil {
 		return err
 	}
+	key := logID.Base64URLString()
+	hash := sha256.Sum256(data)
+	s.stateMu.Lock()
+	if prev, ok := s.lastState[key]; ok && prev == hash {
+		s.stateMu.Unlock()
+		return nil // unchanged; CertSpotter calls this per batch
+	}
+	s.lastState[key] = hash
+	s.stateMu.Unlock()
+
 	_, err = s.pool.Exec(ctx, `
 		INSERT INTO ct_checkpoints (log_id, state, updated_at)
 		VALUES ($1, $2, now())
 		ON CONFLICT (log_id) DO UPDATE SET state = EXCLUDED.state, updated_at = now()`,
-		logID.Base64URLString(), data)
+		key, data)
 	return err
 }
 

internal/store/store.go

@@ -5,6 +5,7 @@ package store
 import (
 	"context"
 	"fmt"
+	"sync"
 	"time"
 
 	"github.com/jackc/pgx/v5"
@@ -20,6 +21,9 @@ import (
 type Store struct {
 	pool    *pgxpool.Pool
 	metrics *metrics.Metrics
+
+	stateMu   sync.Mutex
+	lastState map[string][32]byte // per-log checkpoint hash; skip identical writes
 }
 
 // Open connects to PostgreSQL. The pool is bounded; the caller configures the
@@ -38,7 +42,7 @@ func Open(ctx context.Context, databaseURL string, maxConns int) (*Store, error)
 	if err != nil {
 		return nil, err
 	}
-	return &Store{pool: pool}, nil
+	return &Store{pool: pool, lastState: make(map[string][32]byte)}, nil
 }
 
 func (s *Store) SetMetrics(m *metrics.Metrics) { s.metrics = m }