PRR comments

Signed-off-by: Aldo Culquicondor <acondor@google.com>

keps/sig-apps/2307-job-tracking-wihout-lingering-pods/README.md

@@ -264,7 +264,8 @@ the controller processes all the Pods. The number of Pods is caped by:
 - time: in each iteration, the job controller removes all the Pods' finalizers
   it can in a unit of time in the order of minutes. This allows to throttle the
   number of Job status updates.
-- count: Preventing big writes to etcd.
+- count: Preventing big writes to etcd. We limit the number of UIDs to the order
+  of hundreds. 
 
 If any Pod finalizer removal fails in step 3, the controller manager still
 executes step 4 with the Pods that succeeded.
@@ -307,6 +308,8 @@ the owner reference.
   - Job tracking with feature enabled.
   - Tracking of terminating Pods.
   - Transition from feature enabled to disabled and enabled again.
+  - Tracking Jobs with big number of Pods, making sure the status is eventually
+    consistent.
 - E2E test:
   - Job tracking with feature enabled.
 
@@ -399,7 +402,7 @@ _This section must be completed when targeting alpha to a release._
 
 * **Are there any tests for feature enablement/disablement?**
 
-  Yes, there are integration tests.
+  Yes, we plan to add integration tests.
 
 ### Rollout, Upgrade and Rollback Planning
 
@@ -484,11 +487,13 @@ previous answers based on experience in the field._
 * **Will enabling / using this feature result in any new API calls?**
 
   - PATCH Pods, to remove finalizers.
-    - estimated throughput: one per created Pod, when Pod finishes or is removed.
+    - estimated throughput: one per Pod created by the Job controller, when Pod
+      finishes or is removed.
     - originating component: kube-controller-manager
   - PUT Job status, to keep track of uncounted Pods.
-    - estimated throughput: at least one per Job sync. Extra calls are throttled
-      at 1 per minute.
+    - estimated throughput: at least one per Job sync. The job controller
+      throttles more calls at 1 per a few minutes (precise throughput TBD from
+      experiments).
     - originating component: kube-controller-manager.
 
 * **Will enabling / using this feature result in introducing new API types?**
@@ -507,7 +512,7 @@ the existing API objects?**
     - Estimated increase: new finalizer of 33 bytes.
   - Job status
     - Estimated increase: new array temporarily containing terminated Pod UIDs.
-      The job controller caps the size of the array.
+      The job controller caps the size of the array to less than 20kb.
 
 * **Will enabling / using this feature result in increasing time taken by any 
 operations covered by [existing SLIs/SLOs]?**