Mainframe, MVS and zOS Discussion

There is only one way:  full integration testing using work/test copies of production data and all associated/affected application systems, preferably involving the users of the actual production system(s) and then getting their sign-off on the changes. 

I "got into" mainframes in 1976 and am still working with them today.  Do the math

Check out the IEBPTPCH utility

Example
000001 //EPBPTBCH JOB 'Merging PDS members to PS',CLASS=A,
000002 // MSGLEVEL=(1,1),MSGCLASS=X
000003 //*
000004 //SQLTOSEQ EXEC PGM=IEBPTPCH
000005 //SYSPRINT DD SYSOUT=*
000006 //SYSUT1 DD DISP=SHR,DSN=EP10.JCB.SQL
000007 //SYSUT2 DD DISP=OLD,DSN=EP10.JCB ...

What kind of timeouts are they getting?   AICA abends?  Deadlock abends?   That matters.

That reliability is designed in to both the hardware and software from the ground up. First of all, it's very unusual for a mainframe LPAR to just "go down". Everything about the hardware and software is designed to be resilient, self-recovering, and protect itself against "bad" actors. For example ...

Using symbols is the best solution. Yes, they can get 'busy' but not if you're smart about how you structure them. If the elements you need to vary are LPAR dependent (like in a parallel sysplex environment) then you can control using the 4-byte LPAR system name as part of the reference.
You can ...

Easiest way is to use CURL.

Easy peasy, as long as you don't care about what the specific abend code is. Do this with 2 JCL steps: the first being IDCAMS, the second attempting to execute any non-existing program inside an // IF // ENDIF construct that will run only if the previous IDCAMS step completed with a step return code ...