Android: removes old "keep_test_server_ports".

tools/telemetry/telemetry/core/backends/chrome/android_browser_finder.py

Index: tools/telemetry/telemetry/core/backends/chrome/android_browser_finder.py
diff --git a/tools/telemetry/telemetry/core/backends/chrome/android_browser_finder.py b/tools/telemetry/telemetry/core/backends/chrome/android_browser_finder.py
index f956cedb4bbe931909a7d025754f823a49cfb28a..9686d5b89cb4aae4c8441d60a058463a096c2546 100644
--- a/tools/telemetry/telemetry/core/backends/chrome/android_browser_finder.py
+++ b/tools/telemetry/telemetry/core/backends/chrome/android_browser_finder.py
@@ -198,14 +198,11 @@ def FindAllAvailableBrowsers(finder_options, logging=real_logging):
         if ret:
           logging.warn('Failed to taskset %d (%s)', pid, ret)
 
-    # Experimental device side workaround for crbug.com/268450 (adb instability)
-    # The /sbin/adbd process on the device appears to hang which causes adb to
-    # report that the device is offline. Our working theory is that killing
-    # the process and allowing it to be automatically relaunched will allow us
-    # to run for longer before it hangs.
-    if not finder_options.keep_test_server_ports:
-      # This would break forwarder connections, so we cannot do this if
-      # instructed to keep server ports open.
+    if not os.environ.get('BUILDBOT_BUILDERNAME'):
+      # Killing adbd before running tests has proven to make them less likely to
+      # flake out during the test. We skip this if Telemetry is running under a
+      # buildbot because build/android/test_runner.py wrapper already took care
+      # of it before starting the shards.
       logging.info('Killing adbd on device')
       adb.KillAll('adbd')
       logging.info('Waiting for adbd to restart')