fix(depot): retry transient probe responses like curl --retry

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

internal/depot/remote.go

@@ -11,6 +11,7 @@ import (
 	"os"
 	"path/filepath"
 	"strings"
+	"time"
 )
 
 // NewBackend returns the backend for name ("local"|"cdn"|"bunny").
@@ -85,9 +86,26 @@ func (b *httpBackend) cdnURL(sha string) string {
 	return fmt.Sprintf("%s/%s", b.cfg.CDNBase, BlobKey(sha))
 }
 
+// probeRetrySleep is the backoff sleeper for transient probe retries;
+// tests stub it (same pattern as sweep.go's confirmSleep).
+var probeRetrySleep = time.Sleep
+
 // rangeProbe issues GET <url> with Range: bytes=0-0 and classifies the
-// response. Never HEAD (banned by spec).
+// response. Never HEAD (banned by spec). Transient outcomes (transport
+// error, or a status that is neither 2xx nor 404/410) retry up to 2 more
+// times with linear backoff (1s, 2s) — the CDN edge throttles sustained
+// sweeps; the bash this ports absorbed that with curl --retry 2.
 func (b *httpBackend) rangeProbe(ctx context.Context, url string, headers map[string]string) (ProbeState, bool, error) {
+	for attempt := 0; ; attempt++ {
+		state, transient, err := b.rangeProbeOnce(ctx, url, headers)
+		if !transient || attempt >= 2 {
+			return state, transient, err
+		}
+		probeRetrySleep(time.Duration(attempt+1) * time.Second)
+	}
+}
+
+func (b *httpBackend) rangeProbeOnce(ctx context.Context, url string, headers map[string]string) (ProbeState, bool, error) {
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
 	if err != nil {
 		return Unconfirmed, true, err
@@ -212,6 +230,7 @@ func (b *httpBackend) fetch(ctx context.Context, sha string) (*http.Response, er
 		if resp, err := b.get(ctx, b.cdnURL(sha), nil); err == nil && resp.StatusCode >= 200 && resp.StatusCode < 300 {
 			return resp, nil
 		} else if err == nil {
+			_, _ = io.Copy(io.Discard, resp.Body)
 			resp.Body.Close()
 		}
 		resp, err := b.get(ctx, b.storageURL(sha), map[string]string{"AccessKey": b.cfg.ReadKey})
@@ -219,6 +238,7 @@ func (b *httpBackend) fetch(ctx context.Context, sha string) (*http.Response, er
 			return nil, err
 		}
 		if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+			_, _ = io.Copy(io.Discard, resp.Body)
 			resp.Body.Close()
 			return nil, fmt.Errorf("bunny get %s: unexpected status %d", sha, resp.StatusCode)
 		}
@@ -230,6 +250,7 @@ func (b *httpBackend) fetch(ctx context.Context, sha string) (*http.Response, er
 		return nil, err
 	}
 	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		_, _ = io.Copy(io.Discard, resp.Body)
 		resp.Body.Close()
 		return nil, fmt.Errorf("cdn get %s: unexpected status %d", sha, resp.StatusCode)
 	}