On a bit of a tangent but I might recommend reviewing the following post by my colleague and considering whether a 2 phase purge schedule would make sense for you. One purge task running as standard to clean up the completed messages and one running on a lag to catch any suspended / leftover messages that are old enough that you think they could be removed:

https://community.intersystems.com/post/why-keep-integrity-important-when-purging-healthshareensemble-data

I would generally caution against killing IRIS jobs from outside IRIS, as under rare conditions this could lead to IRIS hanging. If this is a concern I would suggest reaching out to WRC who can help you confirm that a process is safe to terminate.

The journal does not contain the line of code (imagine needing to log that information for every single global modification), so to find the relevant code would require the type of cross referencing I suggested. The other option I would suggest would be to go through the program in the debugger. Especially if you have a general sense of where the relevant change would be made. But again, maybe the best option depends on why you are looking for this specific global set.