The real trick isn’t the staffing, it is the boundary of the systems you need to change. To improve a process in a large system, you almost have to assume none of the other systems you touch have to change anything. Compensate for the large org and large number of critical systems with technical excellence and trickery. Your image upgrade project can’t get any QA resources to sign off on any code changes in the legacy C++ stack? Fix the bug with LD_PRELOAD overrides in your image upgrade program. Take the time to reason very thoroughly about the bug, but go for it. Obviously, use whatever automated validation you have and have a cautious roll out, but whatever you do, don’t be suckered into waiting for a team you don’t absolutely control to do something.