If a username and an email are each sufficient to identify a user then they are the same thing, because every property that can be resolved about one object can be resolved about the other.

Another example: when should you split up one module into two?

Without a clear rule of thumb you pick some arbitrary point between every function having its own file and every function being in one file.

Your code should be organized around the information that identifies it.

The better answer for having two microservices was that they have different SLOs but that usually leads to fewer services.

When micro services were trending, answering this question was super difficult.

The answer was vaguely: we want two microservices when we want to deploy separately, but why do you want to deploy two applications?

The answer was usually that there were two teams? Well why two teams?

Do you have a rule of thumb for determining if a unit of behavior is actually two separate units of behavior?

The best way to learn any skill is to get feedback quickly and frequently.

Software testing is a hack for frequent feedback.

Not only is software testing a limitation for skill development, it's a super power when used correctly.

Copying code from the Internet? Test it to confirm it works

The engineer should take ownership for breaking prod.

The manager should take ownership for an engineering culture that convinced an engineer to hide their mistake.

There's only one way to get the culture you want: demonstrate it every chance you get

filtered = Q(l, name__contains="k", age__lt=20)

Could be written as
filter(lambda x: "k" in x["name"] and x["age"] < 20)

I could imagine the former has the benefit of checking for the value existing in the dict but otherwise I think this is better solved without a package