I'm not sure how it got to this state but I found that when I saved a copy of one particular strategy and ran that copy and I would get consistently completely different results on the same dataset over the same parameters.

Confused at this, I thought maybe it's because I installed some extensions recently that I messed this up.. so I uninstalled some. And now when I run the same strategy on the same dataset, I get results like:

All Trades Long Trades Short Trades Buy & Hold
Starting Capital $20,000.00 $20,000.00 $20,000.00 $20,000.00
Ending Capital $7,018,109,414,048,540,000,000,000,000.00

Note that huge number for ending capital. The expected value for this backtest should have been something in the tens of thousand still since the test was run over 3 years. Anyone experienced similar problems before and know the fix?

This is flattering to know when someone thinks we can offer a fix for unknown code running on unknown data by unknown data provider with mostly unknown settings (except for the data loading range) with unknown strategy parameters (if they exist). Sorry to disappoint you but we need to collect some required information first:

How to report a Problem? - especially the Suggestions and Summary sections.

Note: uninstalling/reinstalling extensions and/or WL6 is a harmless but useless step that is never going to solve any problems like that (wish it was that simple). With a commission addin as the exception, existing extensions hardly could affect backtest results (especially in this manner).