For info, the actual program in which this logic appears does this:

Now that would be interesting to see...

But for this specific problem, I modified your thread code as follows:

sub RunThread {
    my $tid = threads->tid;
    my $dbh = DBI->connect("DBD::Pg",...) or die "failed connect";
    my ($x, $y)=@_;
    open (my $fh2, ">/tmp/da-$x-$y") or die "failed in thread $! at it
+er $x thread $y " ;
    print "$tid : ", fileno( $fh2 ); sleep 1;
    close ($fh2);
    $dbh->disconnect or die "disconnect failed";
}
[download]

When I run that, I get the following which shows that the filenos underlaying the file handles are being reused by each new batch of threads. If you don't see similar, then it probably means that Solaris/Perl on your system is downlevel or broken.

[19:04:06.98] c:\test>DBjunk.pl
1 : 3
2 : 4
3 : 5
4 : 6
5 : 7
6 : 8
8 : 9
10 : 10
9 : 11
7 : 12
12 : 13
14 : 14
13 : 15
11 : 16
15 : 17
16 : 3
17 : 4
19 : 5
18 : 6
20 : 7
21 : 8
22 : 9
23 : 10
24 : 11
25 : 12
27 : 3
28 : 4
26 : 5
29 : 6
30 : 7
31 : 8
32 : 3
33 : 4
34 : 5
35 : 6
36 : 7
37 : 8
38 : 9
39 : 10
40 : 11
41 : 12
[download]

Thanks, no I don't see that - the file numbers gradually increase. See discussion below - looks like Oracle client is leaking :(
Thanks for the replies, a very fruitful introduction to perlmonks!

Maybe your open filehandles get cloned for every thread? You could instead first create your threads and then later on allocate the resources only needed in one thread.