Massive Non-DB paging on (nearly) quiescent system

So I've got a machine whose only official purpose in life is as an offsite HA machine. It talks to our production machine over an adequate but not extremely large pipe. It's very simply configured with three pools, machine, base and spool.  It's running at 8% CPU utilization, and that's nearly all due to the QEDD (replication) jobs (and an occasional appearance by our good friend QDBFSTCCOL).

And yet, I'm getting a ridiculous amount of Non-DB faulting and paging.  I mean it's crazy:

% CPU used . . . . . . . :        7.8    System ASP . . . . . . . :     9674 G
Elapsed time . . . . . . :   00:09:20    % system ASP used  . . . :    32.9795
Jobs in system . . . . . :        819    Total aux stg  . . . . . :     9674 G
% perm addresses . . . . :       .108    Current temporary used . :    21897 M
% temp addresses . . . . :      1.118    Peak temporary used  . . :    25593 M


Sys      Pool   Reserved    Max  ------DB------- ----Non-DB-----
Pool    Size M   Size M     Act  Faults    Pages Faults    Pages
  1    3110.39   1519.69  +++++       .0      .0 .0      .0
  2   52876.82      9.46    372     14.3  1583.1 623.8  2405.6
  3    6220.78       .00    126       .0      .0 .0      .1

It's been a while since I tracked this stuff down.  Where do I go for an initial look?  And WRKSYSACT shows nothing out of the ordinary either.  The QEDD jobs, and tiny percentages from things like RMTMSAFETA and SMXCAGERnn.

Thanks as always!