Tim, I don't know what I'm missing but running single tests always fail with this option. Here is the command I use:

autoninja -C out/Debug blink_tests &&  ./third_party/WebKit/Tools/Scripts/run-webkit-tests --no-retry-failures --additional-driver-flag=enable-threaded-compositing --debug  external/wpt/pointerevents/pointerevent_attributes_hoverable_pointers-manual.html

Am I missing something? But when I run a bunch then some of the tests pass. I don't think this is a problem with pointerevents tests per se.

Robert, any ideas on why these would be failing like this?

It seems to me when --additional-driver-flags=--enable-threaded-compositing is added, the first test on each content_shell process almost always fails. If you add --repeat-each=10 and run only one test (that's known to pass even with threaded compositing), you'll see the first run fails and all the remaining pass.

The first run seems to always print out a layout tree which is actually a directory list of the root (/) on my workstation. Something is wrong with the first navigation/loading when threaded compositing is enabled?

I ran all pointerevent tests listed in this bug one by one with --repeat-each=10 and --additional-driver-flags=enable-threaded-compositing and for all of them the first one failed and the rest passed. To be accurate we have changed some of the stuff since this bug was filed for the pointerevents. Those could have fixed this issue. So I don't think at this point there is any specific issues with pointerevents and threaded compositing.
I assume you might want to track the first test failure in a separate bug. I'll close this for now. Let me know.

These are failing because the layout isn't stable. I have a fix for them

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f8019d260dd3215d5b303ec5dec1da9121e70034

commit f8019d260dd3215d5b303ec5dec1da9121e70034
Author: Dave Tapuska <dtapuska@chromium.org>
Date: Fri Sep 14 15:19:43 2018

Amend pointerevent tests so they are stable while the test is executing

Some pointer event tests used a log method that would cause the layout
to change and this caused issues for using threaded compositing.

Switch to a method that already existed before to collect the log entries
and then show them at the end of the test.

BUG= 831653 

Change-Id: I3d1b16d01aa15c3c1660d36ce10aa702671c6786
Reviewed-on: https://chromium-review.googlesource.com/1226097
Reviewed-by: Navid Zolghadr <nzolghadr@chromium.org>
Commit-Queue: Dave Tapuska <dtapuska@chromium.org>
Cr-Commit-Position: refs/heads/master@{#591349}
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_capture_mouse-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_capture_suppressing_mouse-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_lostpointercapture_for_disconnected_node-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_lostpointercapture_is_first-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_releasepointercapture_onpointercancel_touch-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_releasepointercapture_onpointerup_mouse-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_setpointercapture_relatedtarget-manual.html
[modify] https://crrev.com/f8019d260dd3215d5b303ec5dec1da9121e70034/third_party/WebKit/LayoutTests/external/wpt/pointerevents/pointerevent_support.js