Is 3.6m the build time of the remained Java code, the rewritten Go code, or both remained Java and rewritten Go code together?
Also doable in Java.
So 12 months of effort -> at Bay area salaries and costings that is in the order of magnitude of 300,000USD. Your parent company had 6.6 million NZD profit last financial year. So for investing around 5% of the total profit of the larger group your mayor benefit is reduced compile times. Plus some other second system benefits.
That is a rather expensive decision to have taken, and I think if you had been forced to do a proper ROI investigation before hand you would not have gotten that. In my experience when doing these calculations and getting the numbers down it is never the financially wise choice to change languages but instead actually fix the pain points that people have with the language and most often the actual project setup. It is only when a language change is incremental, adds a key feature or is a market requirement e.g. for deploying on certain devices that a new language makes financial sense.
I am not judging your GO decision over Scala as a better language but I am judging your management layer for doing this. Especially as it sounds like from other posts that GO is now 1 more language to support in your company. Increasing certain costs outside of yours as a team in cross team training opportunity, support and monitoring etc...
Of course sometimes changing language is like changing from speaking French to Italian in the office and hoping that office politics will disappear... Been there done that, it did not work out as hoped by the developers.
Then it was a Java to GO move. I wonder where your compiles take the most time. For our work building the jar files are the most expensive part. Javac is about 20 seconds all in for about 3000 files to compile on my 2009 macbook.