I got excited about this and decided to run some tests.  I ran the hold targeter 4 times in a row (with a 1-second recheck interval to force updates) both pre and post patch.  Much to my surprise, the delayed logging mechanism actually made things very slightly slower :\

-- pre-patch timing at WARN loglevel                                           
                                                                               
real    0m23.509s                                                              
user    0m0.309s                                                               
sys     0m0.028s                                                               
                                                                               
real    0m23.737s                                                              
user    0m0.304s                                                               
sys     0m0.041s                                                               
                                                                               
real    0m23.825s                                                              
user    0m0.313s                                                               
sys     0m0.020s                                                               
                                                                               
real    0m22.847s                                                              
user    0m0.310s                                                               
sys     0m0.032s                                                               
                                                                               
avg: 23.4795 

-- post-patch timing at WARN loglevel                                          
                                                                               
real    0m25.774s                                                              
user    0m0.304s                                                               
sys     0m0.037s                                                               
                                                                               
real    0m24.530s                                                              
user    0m0.332s                                                               
sys     0m0.020s                                                               
                                                                               
real    0m22.944s                                                              
user    0m0.322s                                                               
sys     0m0.053s                                                               
                                                                               
real    0m24.090s                                                              
user    0m0.309s                                                               
sys     0m0.033s                                                               
                                                                               
avg: 24.3345 

I also tested at INFO level with similar results.

Crazy!  Additional sub tracking, more runtime call stacks, gremlins, etc.  In any event, I posted a rebased branch for my testing.  It includes a signoff, but I can't at this point see that the code changes are worth it or even a good idea.  Other types of testing might show otherwise, though.


https://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/user/berick/lp1840053-log-subs