X crashes after 'Resetting rcs0 for preemption time out'

A new system being enabled has X crashing after idling for a while, or more easily reproduced by changing the screen configuration with xrandr. dmesg with a recent drm-tip build is attached. dmesg_20191126.log

It took over 10s, looks to be a genuine hang. E.g. the final death throes is

[   77.074520] Asynchronous wait on fence i915:gnome-shell[1732]:5de timed out (hint:intel_atomic_commit_ready+0x0/0x54 [i915])
[   79.539492] i915 0000:00:02.0: Resetting rcs0 for preemption time out
[   79.539534] i915 0000:00:02.0: Xorg[1600] context reset due to GPU hang
[   79.539612] [drm:__i915_request_reset.cold [i915]] context Xorg[1600]: guilty 3, banned
[   79.539687] [drm:__i915_request_reset.cold [i915]] client Xorg[1600]: gained 4 ban score, now 6

I was going to suggest trying the 640ms timeout, but it's dead Jim.

Usual question: confirmed present in drm-tip?

it was built from cod/tip/drm-tip/2019-11-27 1815badcd921b2cc41ee5ef2fc41c26b3d3ebbf4

mentioned in issue #166

mentioned in issue #201 (closed)

mentioned in issue #97 (closed)

mentioned in issue #219 (closed)

mentioned in issue #122 (closed)

mentioned in issue #125 (closed)

mentioned in issue #203 (closed)

mentioned in issue #203 (closed)

mentioned in issue #203 (closed)

mentioned in issue #264 (closed)

mentioned in issue #311 (closed)

mentioned in issue #353 (closed)

mentioned in issue #491 (closed)

mentioned in issue #515 (closed)

mentioned in issue #520 (closed)

mentioned in issue #520 (closed)

mentioned in issue #526 (closed)