I am not a fan of it, as I really like when my script knows how many tests it is going to run, but sometimes it is really difficult to know up-front. It might depend on the number of files in a directory. We might be able to compute the number at the beginning of our test-script, but in some cases this might be too difficult.

In that case we can test without a plan or we can call done_testing.

use strict;
use warnings;

use lib 'lib';
use MyTools;

use Test::More;

my $x = sum(1, 1);
is $x, 2,  '1+1';

my $y = sum(2, 2);
is $y, 4,  '2+2';

done_testing;

The output of such test-script is quite similar to the case when we used no_plan.

ok 1 - 1+1
ok 2 - 2+2
1..2

The difference is that in this case we explicitly need to tell the testing environment that we have finished all the tests. This means if we exit() early, even the line immediately before done_testing() we won't get the plan printed at the end, instead we'll have a comment:

$ perl test.t

ok 1 - 1+1
ok 2 - 2+2
# Tests were run but no plan was declared and done_testing() was not seen.

If we use the harness our test will be reported as failed:

$ prove test.t

t/test.t .. 1/? # Tests were run but no plan was declared and done_testing() was not seen.
t/test.t .. Dubious, test returned 254 (wstat 65024, 0xfe00)
All 2 subtests passed 

Test Summary Report
-------------------
t/test.t (Wstat: 65024 Tests: 2 Failed: 0)
  Non-zero exit status: 254
  Parse errors: No plan found in TAP output
Files=1, Tests=2,  0 wallclock secs ( 0.04 usr  0.01 sys +  0.11 cusr  0.01 csys =  0.17 CPU)
Result: FAIL

We will still disregard cases when a loop that was supposed to run 1000 times, testing 1000 things, ran only 3 times, but it is a step up from no_plan.

Please don't overuse it.