When does Save While Active (SWA) (with BRMS) checking pointing begin?

Hoping a backup/recovery (BRMS?) guru out there can shed a little
light on what's going on here..

--------From the joblog (job name SV_PROD)------------------
23:00:00 STRBKUBRM CTLGRP(SV_PROD)
23:00:29 Starting save of library HLTHPRDFIL
23:00:29 Starting save of library HLTHPRDPGM
23:00:29 Starting save of library PRCPRDFIL
<others snipped>
03:10:24 3519 objects saved from library HLTHPRDFIL
03:10:24 4967 objects saved from library HLTHPRDPGM
03:10:25 506 objects saved from library PRCPRDFIL
<others snipped>


-------From QHST (and SWA msgq )-------------------
23:01:22 Save-while-active checkpoint processing for library HLTHPRDFIL complete
02:27:23 Save-while-active checkpoint processing for library
HLTHPRDPGM complete.
02:28:21 3519 objects saved from library HLTHPRDFIL.
02:29:22 Save-while-active checkpoint processing for library PRCPRDFIL complete.
02:29:25 4967 objects saved from library HLTHPRDPGM.
02:25:52 Save-while-active checkpoint processing for library <snip> complete.
02:35:53 506 objects saved from library PRCPRDFIL.
<others snipped>
03:13:46 BRMS backup job SV_PROD ended on S10316ZM

The libraries in the BRMS group are all defined with SWA = "By
Object". Which as I understand it is the same as *SYSDFN on the
SAVACT parameter of the SAVLIB command. (Yes, I know that *SYSDFN is
a bad idea and I've pointed it out to management)

My question...
When does checkpoint processing begin for the 2nd library in the
group, HLTHPRDPGM?

Looking at the above, I would say that the answer is (basically) after
the objects are in saved 1st library, HLTHPRDFIL.

However, at 02:02 we had an RPG program attempt to use a data area in
HLTHPRDPGM and not be able to because it was locked. The RPG program
in question is the only one that would access the data area, so I can
only assume that the data area was locked because the checking
pointing of HLTHPRDPGM hadn't completed which would lead me to
conclude that checkpoint of all the libraries started at 23:00. But
then why, especially with SWA = *SYSDFN, would checkingpointing not
finish on HLTHPRDPGM till 02:27:23 when most of the objects in it are
*PGM, DSPF, PRTF, and some *DTAARA?

Thanks!
Charles Wilt