It seems to me that there's some issue in the Orchestrator backstage plugin backend.
I wrote a simple nodejs script that reproduces the problem (included in slack thread).
The script runs a POST request that updates the description of the workflow with a random number, waits for the result and prints the description. It follows that with a GET request, waiting for result and printing the description.
There are two possible options: Either the GET request returns right away with the wrong description (the previous one),
or it waits about half a minute and then returns the correct result (the updated description).
To reproduce it, I ran in one terminal: yarn start:backstage
And in another terminal: node <script location>

https://redhat-internal.slack.com/archives/C059YMMB664/p1702217587624919