Debugging is a major weak point for mathematica, especially for code of any size. Mathematica is very very good at some things, but i am building something medium size and growing and debugging and a few other things made it very difficult and slow to develop. It turned out the one very cool thing Mathematica did for me was handled fine by numerical integration in python using numpy and scipy, and some calculations are actually handled better. With some effort I moved my code to python and will stay there. I will use Mathematica for one-off stuff, or to double-check a few results. I wish I had that advice 9 months ago.

I never got WorkBench (2025) to work with Eclipse on windows. I was tempted to try to write a debugger (I've done that for another platform) but didn't get very far with the research. There isn't a documented interface to the kernel, and I needed to get my actual task done.

Print statements are a very slow and inefficient way to debug. Trace and echo are fancy print statements. There is a debugger for notebooks, but it is very unintuitive and not very powerful. I built a simple system to log messages to a file, but another set of print statements. At least there you can externally filter and import messages without selectively turning things off and on trying to narrow something down. Ive built a more sophisticated logging system with tags on logging statements no another platform, with a way to selectively log tags. That system is good for troubleshoot things in a customer's production environment where you can't debug, but not a great debug tool

Is WorkBench worth another try?