The perl script that runs 24/7 does this:

Net::DBus::Reactor->main->add_timeout (50, Net::DBus::Callback->new(method => \&Utils::Monitor::do_monitor_files));

This results in 20 wakeups per second.

It's particularly humourously bad because of this:

sub do_monitor_files
{ 
  my ($event, $data, $func, $path, $object);

  return if (!$has_gamin);

....


I don't even have Sys::Gamin on my system.

The real fix for this problem is to get the fd to the gamin socket and register it with Reactor->main->add_read().

This seems to be somewhat impossible, though, as the Sys::Gamin code I found (on CPAN) doesn't let you know the fd.

A workaround for now might be to do these three things:
  (a) massively scale back the poll rate
  (b) only poll when changes in the files will actually matter
  (c) don't poll at all if Sys::Gamin isn't even installed