Another test script, which might be even more relevant:

  trait Foo {public $var = [];}
  class Bar {use Foo; public $var = [];}

This also causes a fatal error for me, which should only be a strict error based 
on the documentation.

I'd call this a bug, rather than a documentation issue; if Foo::$var === Bar::$var, I can't really see why it should fatal.

Reclassifying, and we'll see what the engine folk think.

If it is logically the same thing, it should be only in one of these traits, don't 
you think?
SRP and DRY, fellas

This is a bug. It should not a be a fatal error if the initial value is even the 
same **type**. Consider using traits to define default static vars. Defining them 
through inheritance, these static vars are clobbered unless explicitly redefined 
in every child class.

It makes perfect sense not to allow this with constants, but why static 
variables?

I believe this is going to be problematic, because the object/array is not the 
same, i.e., it is not the same literal, and I assume that I currently check for 
object identity. Object identity works for most literals, but not arrays.

To fix this, special care is required. A value comparison might be possible, and 
would be necessary here. A simple check for the type is not sufficient, because 
$var = 1; vs $var = 2; should result in the expected error.

But yes, in general, I consider this behavior a bug.

In PHP7.0, this E_STRICT was dropped as described in the 7.0 docs: http://php.net/manual/en/migration70.incompatible.php

> Same (compatible) property in two used traits	- Notice removed, triggers no error

However, the documentation on traits still mentions the E_STRICT.

The documentation is indeed out of date.

No E_STRICT in PHP 7:
- https://3v4l.org/Jafoi matching in two traits
- https://3v4l.org/pQYTl matching in traits and class

Conflicting definitions:
- https://3v4l.org/brZt0 conflict between traits
- https://3v4l.org/umfVE conflict between trait and class

I've included a patch for the documentation on this ticket as well.

The following snippet results in the "definition differs and is considered incompatible" fatal error in 7.0+. Can this be considered as a bug? 

const VALUE = true;

trait Foo {public $var = VALUE;}
trait Bar {public $var = VALUE;}
class Baz {use Foo, Bar;}

https://3v4l.org/kkHFe

Closing as docs have been fixed.