perl-Net-Statsd-Server-0.17-1.fc22 can fails to build like this:

2014-07-18 16:52:50.000000 +0200 note  Net::Statsd::Server: Initializing Net::Statsd::Server::Backend::Console backend
2014-07-18 16:52:50.000000 +0200 note  Net::Statsd::Server: Initializing Net::Statsd::Server::Backend::Graphite backend
2014-07-18 16:52:50.000000 +0200 info  AnyEvent: Autoloaded model 'AnyEvent::Impl::Perl', using it.
2014-07-18 16:52:50.319436 +0200 note  Net::Statsd::Server: statsd server started on 0.0.0.0:40001 (v0.17)
2014-07-18 16:52:50.319499 +0200 note  Net::Statsd::Server: manager interface started on 0.0.0.0:40002
2014-07-18 16:52:50.319549 +0200 note  Net::Statsd::Server: metrics flush will happen every 1s
2014-07-18 16:52:51.319011 +0200 note  Net::Statsd::Server: flushing metrics
Flushing stats at Fri Jul 18 16:52:51 2014
{"timer_data":{},"pctThreshold":[90],"timers":{},"sets":null,"counters":{"statsd.packets_received":0,"statsd.bad_lines_seen":0},"gauges":{},"counter_rates":{"statsd.packets_received":0
,"statsd.bad_lines_seen":0}}
#   Failed test 'Backend recognised our bad message bumping 'bad_lines_seen''
#   at t/integration-tests/bad-message.t line 31.
#          got: '0'
#     expected: '1'
# Looks like you failed 1 test of 4.
t/integration-tests/bad-message.t .......... 
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/4 subtests 

This is a race in tests (or Test::Statsd module) and can be reproduced easily by inserting "sleep 10;" into t/integration-tests/bad-message.t between send_udp() and wait_and_collect_flush_data() invocations:

$t->start_statsd();
$t->send_udp(localhost=>40001, "a_bad_test_value|z");
sleep 10;
# Will wait until it receives the graphite flush
my $stats = $t->wait_and_collect_flush_data();
$t->stop_statsd();

Such test is unreliable and fails randomly, especially on a loaded host.