What kind of constraints do you have? Does each process needs to get the same amount of bytes? The same amount of lines? Or if you have N processes, will approx. 1/N of the lines do? Can the lines be fed in the same order as in the file, or does that have to be random as well? Can you just chop the file into N equal pieces and shuffle those? If not, why not? Would the following algorithm do?

1. Let the number of lines in the file be L. Let the number of processes be N. Let A_i = 0 for 0 <= i < N. Label the processes P__i, 0 <= i < N.
2. For each line l, 0 <= l < L, assign this line to process P__i with chance (L/N - A__i)/(L - l). For the process P__j which is picked this way, increment A__j.
3. If necessary, shuffle the lines assigned to each process.

Some pseudo, untested, code:

my $L = ... number of lines ....;
my $N = ... number of processes ....;

# Open filehandles for each process:
my @fh;
for (my $i = 0; $i < $N; $i ++) {
    open $fh[$i], "| $process" or die;
}
    
# Initialize @A:
my @A = (0) x $N;
     
my $l = 0;
    
# Iterate over the input.
while (<$input>) {
    # Pick a number.  
    my $r = rand($L - $l);

    # Find process.
    my $i = 0;
    while ($r >= ($L / $N - $A[$i])) {
        $r -= $L / $N - $A[$i];
        $i ++;
    }  

    # Write line.
    print $fh[$i] $_;
    
    # Increment array.
    $A[$i]++;
     
    # Increment line counter;
    $l++
}
[download]