You asked about an alternate approach. I have often found that data like you have is best parsed line by line instead of being slurped into a single variable. Of course your mileage will definitely vary!

Anyway a "read the line and throw it away" if not needed is much faster than a "slurp" and substitution. A simple formulation of this is shown below.

Update: I would also add Flipin good, or a total flop? as another way along the way of shem's approach. The three dot (...) version of the "flip flop" syntax works with multiple lines, the (..) version works with single lines.

#!/usr/bin/perl -w
use strict;

my $line;

while ($line = <DATA>)
{
   skip_record() if $line =~ m/^\s*CD_TEXT/;
   print $line;
}
sub skip_record
{
  while ($line = (<DATA>), $line !~ m|^\s*//\s*Track|){};
}

#prints: 
#  CD_DA
#
#  // Track 1


__DATA__
CD_DA

CD_TEXT {
  LANGUAGE_MAP {
    0: 9
  }
  LANGUAGE 0 {
    TITLE "Multi-01"
    PERFORMER ""
    SIZE_INFO { 1,  1, 19,  0,  3,  2,  0,  0,  0,  0,  0,  0,
                0,  0,  0,  0,  0,  0,  0,  3,  7,  0,  0,  0,
                0,  0,  0,  0,  9,  0,  0,  0,  0,  0,  0,  0}
  }
}

// Track 1
[download]