[GCM] Update backoff policy to be more conservative.

google_apis/gcm/gcm_client_impl.cc

Index: google_apis/gcm/gcm_client_impl.cc
diff --git a/google_apis/gcm/gcm_client_impl.cc b/google_apis/gcm/gcm_client_impl.cc
index 14a94687a923d5672cf583caf1aefd8a94183a86..3100e71e029d677c85e8d448ff43d38af437c7c5 100644
--- a/google_apis/gcm/gcm_client_impl.cc
+++ b/google_apis/gcm/gcm_client_impl.cc
@@ -27,8 +27,10 @@ namespace gcm {
 
 namespace {
 
-// Backoff policy. Shared across reconnection logic and checkin/registration
+// Backoff policy. Shared across reconnection logic and checkin/(un)registration
 // retries.
+// Note: In order to ensure a minimum of 20 seconds between server errors (for
+// server reasons), we have a 30s +- 10s (33%) jitter initial backoff.
 // TODO(zea): consider sharing/synchronizing the scheduling of backoff retries
 // themselves.
 const net::BackoffEntry::Policy kDefaultBackoffPolicy = {
@@ -37,17 +39,17 @@ const net::BackoffEntry::Policy kDefaultBackoffPolicy = {
   0,
 
   // Initial delay for exponential back-off in ms.
-  15000,  // 15 seconds.
+  30 * 1000,  // 30 seconds.
 
   // Factor by which the waiting time will be multiplied.
   2,
 
   // Fuzzing percentage. ex: 10% will spread requests randomly
   // between 90%-100% of the calculated time.
-  0.5,  // 50%.
+  0.33,  // 33%.
 
   // Maximum amount of time we are willing to delay our request in ms.
-  1000 * 60 * 5, // 5 minutes.
+  10 * 60 * 1000, // 10 minutes.
 
   // Time to keep an entry from being discarded even when it
   // has no significant state, -1 to never discard.