Passing variables to a subroutine on sequence match

Consider the following concurrent assertion

  bit clk ;  
  always #5 clk = !clk ;
  int a = 4 ;    
  always @(posedge clk) a <= a + 1 ;
  task automatic sub1( int ip1 , ref int ip2 );          
    $display("ip1 == %0d",ip1);
    $display("ip2 == %0d",ip2);
    $display("a == %0d\n",a);
    property p;       
      @(posedge clk) (1,sub1(a,a)); 
    p1:assert property(p);
    initial  #26 $finish();

LRM says “The subroutines are scheduled in the Reactive region, like an action block.

[Q1] Shouldn’t we observe updated value of ‘a’ in all 3 $display messages as the task is called in reactive region ?

[Q2] What should the ideal output be ? There is discrepancy with eda tools for message “ip2 == …”

Ref: Edit code - EDA Playground

task automatic sub1( int ip1 , ref int ip2 );          
    $display("ip1 == %0d",ip1);
    $display("ip2 == %0d",ip2);
    $display("a == %0d\n",a);
 ...     @(posedge clk) (1,sub1(a,a)); 
 1800'2023 13.5.2 Pass by reference 
 Arguments passed by reference are not copied into the subroutine area, 
 rather, a reference to the original argument is passed to the subroutine.
 When the argument is passed by reference, both the caller and the subroutine 
 share the same representation of the argument; therefore, any changes made to the argument, 
 within either the caller or the subroutine, shall be visible to each other. 
 The semantics of assignments to variables passed by reference is 
 that changes are seen outside the subroutine immediately (before the subroutine returns). 

[Ben] Based on this, I can see why ip2==5 is correct since the "a" being passed 
is a reference to the variable "a" which is now 5. 

The correct result is ip1 == 4  // the sampled value
ip2 == 5   // the variable "a" value at call time 
a == 5     // the value in the Observed region

Another opinion on this: This is a gray area in the LRM. The arguments passed to the task should be sampled, but if it is passed by reference, it is not clear what applies. Clearly one vendor decided to sample it too, not the reference but the variable referred to.

As far as ‘a’ is concerned, in principle it should be sampled.
So ip1 should be 4 ip2 is debatable, 4 or 5…
It anyway does not make sense to pass by ref in the case of assertions.