What's the best thing to do when you've run into a debugging dead end?

Turn off the computer and go for a walk.

Make a cup of coffee > Explain the problem to someone else > Have a mid-sentence brainwave > Run away back to desk and fix in seconds > Have existential crisis about why that took so long

I think this summarizes my usual process for debugging tricky problems:

• When working through various scenarios, make sure to carefully document each step. Make sure to alter only one thing at a time and check all possible combinations of things that may be causing the problem.

• Break the problem all the way down to its fundamentals - sometimes that may mean starting from scratch and adding code bit by bit until the problem manifests itself.

• Take a break: It's amazing how new ideas can come along when you're taking a walk or are in the shower!

• Research: Look for answers online

• Ask for help: Ask someone who may be able to help, either online or in person

• Workaround: Is there some way to solve the problem without attacking it directly? Maybe there is a way to solve a slightly different problem that will still produce the desired results, or at least get close enough to get past the current bottleneck.

Start asking questions: Why am I doing this? Am I happy? I can't solve this, I'm an impostor. I better work with wood or open a bar..
Oh, that was a typo! Great, keep coding

Ask someone else too look at the problem with you. Sometimes they will see something you didn't since they don't have a ton of assumptions on how it will work.

Stop locally and restart from the simplest working context.
It's almost like performing a git bisect.

0: disable/comment almost everything so that you have the simplest context;
1: test if the context behaves like it's supposed to behave;
2: if it does then re-enable/uncomment half of the things back and GOTO :1;
3: if it does not, then if the problem is self-evident, fix it and RETURN 🥳;
4: if it does not, you already know that the things just restored are responsible for the issue. You can pinpoint the issue even further by disabling/commenting half of the things you have just restored and GOTO :1;

The key point here is that you waste a few minutes in order to get to :0 but avoid wasting several hours finding a bug that is probably way above the context you're looking at.

It can appear vague but it's a debugging process that can be performed in a wide variety of contexts, not just source code.

I always start with asking myself "What assumptions am I making about what's happening?" Most of the time this leads to me realizing I don't understand how something is actually working and that my assumptions created some false understanding. Bugs on computers are almost always deterministic at small scales. So, if you can dive far enough down the abstractions you can find the bug.

I highly recommend a cup of jasmine tea w/honey & meditation sess.

If you dont want to leave your train of thoughts by leaving the desk/computer (yet):
Use the help provided by rubber duck (can also be a text editor or colleague if you prefer):
Explain everything you did so far, as you would need to in a stack overflow question or thorough bug report.
In most cases I do that, the missing piece get's obvious.
If you have the patience to write it down and it doesn't help you can of course ask that question here or somewhere else, or file that bug...