I realized, that this command gives a NULL-pointer as soon as a notebook gets closed:

ALLEGRO_LOCKED_REGION* ar =
al_lock_bitmap(al_get_backbuffer(al5_display), ALLEGRO_PIXEL_FORMAT_ANY, 0);

Just looping until a pointer is available again isn't possible.
I then get an exception at d3d_bmp.cpp, line 767:

if (d3d_disp->render_target->LockRect(&d3d_bmp->locked_rect, &rect, Flags) != D3D_OK) {
ALLEGRO_ERROR("LockRect failed in d3d_lock_region.\n");
return NULL;
}

I need 'ar' for fast pixel draeings. Is there a solution?

It's a self-compiled allegro 5.08 - maybe due to my compilation?
DirectX9c used.

It's a directx "feature." You need to use al_d3d_set_restore|release_callback. Google for "Direct3D lost device" or similar for background.

Isn't there a display lost or a switch out event when the notebook gets closed? If so, you can just set a flag and not draw during that period.

An event-message is too critical. This also can happen between the event-request and the drawing. I have to try Trent's hint.

It's my first trial with 5.x. There was never a problem using Allegro 4.2.
Miss the midi-driver too and I hate it that the event-queue gets flooded when you are at a breakpoint.
Just had to port it because a friend has problems with Allegro 4.2 on a 64-bit-system. This all ain't no fun anymore. Think I'm getting old...

Think it's worth to post the solution then, because this can happen to almost each program that uses Allegro 5. Will try my very best.

I fixed a couple of bugs with respect to device lost handling on the 5.1 branch. These changes are not yet in any released version. 5.1.5 will be released soon.

Do you really need them in the common case? The documentation only mentions "d3d resources you have created yourself". AFAICS, the mechanisms in 5.0 are enough to handle device lost otherwise.