Jobs interchangeably running on different servers in a cluster

I built a Denormalization transformation and added it to the Scheduler. Sometimes, it works great. It often runs successfully but only on the 1st 300 records in my Excel Data source. In the Scheduler Monitor, I see a line of “This is a trial version and is limited to 300 records per transfer”.
I don’t get this message every time it runs in the Scheduler, but I have not been able to find a pattern. Any idea what’s causing the issue?
I’ve tried to narrow down the problem, I think it’s in the Scheduler. The Scheduler in Production is using the server in QA sometimes when it runs a job.
Is there any way to disconnect the QA Server from the Scheduler?

Can you please confirm if both the Production and the QA server are in the same cluster?
Also, confirm if ‘Any Server’ is specified in the Server of the specific schedule? (Refer to the SS below)

We suspect that both the Production and QA Server are pointing to the same repository and if you have specified the schedule to run on any server, it inter-changeably runs on the Production and QA server, depending on the availability of servers.
To avoid this issue, we recommend you specify the Production Server in the ‘Server’ option for the specific schedule.

Very curious! I don’t have that dropdown option in my menu. Instead, there’s a checkbox to “Run this and its child jobs on the same server.” I’m on version 8.4, do I need to update?

Yes, we introduced this option in v8.5.
Can you please confirm if both the production and QA server uses the same repository? If so, this explains why you are encountering this behavior. As it is running the job on any available server.

Thanks. I’m sure we’re pointing to the same repository, we’ll separate them.
As for the upgrade, is there any reason not to skip v8.5.4.623 and go straight to v9?

You can upgrade to v9 as well, but we recommend migrating your flows from v8 after testing them in the QA environment.
Whereas, upgrading to 8.5 should be simpler and straightforward.