Instead of split on + and rejoin with space, you could use a regex for that. The split already has some regex overhead.

use warnings;
use strict;

my @tests = ("this     is+   +test",
             "this+test",
             "morepluses+++++++test");

for (@tests)
{
   s/\++/ /g; #multiple+ to single space (see note)
   print "$_\n";

}
__END__
this     is     test
this test
morepluses test

or could use this regex to also compress spaces:

s/[\s\+]+/ /g; #multiple spaces or + to single space
this is test
this test
morepluses test
[download]

Note: in /\++/ the first + is escaped to mean a literal + sign, the second plus is a "quantifier command" to the regex engine, meaning "one or more". Your split error arises because of this second meaning and there is nothing before the + to "quantify".

Update: If you just want to convert each + to a space. tr will run much faster at that job because it builds a simple translation table. It does not use a regex, so the + should not be escaped:

use warnings;
use strict;

my @tests = ("this     is+   +test",
             "this+test",
             "morepluses+++++++test");

for (@tests)
{
   tr /+/ /;  #lower overhead than the regex engine
   print "$_\n";

}
__END__
this     is     test
this test
morepluses       test
[download]