Right, the end-handling of blocking the intersection is different by necessity (just imagine beyond the end is a cliff that cars fall off so the exit from the intersection is never blocked!). Anyway, the rules for those scenarios are covered by the spec:

• It is illegal to queue across an intersection. In other words, DO NOT enter an intersection unless you can see there is safe passage through it and out the other side!
• If there is a traffic light at the end of the road then its exit is never blocked

I may agree that a gap emerge when two cars pass green.
But it's strange, the same thing don't happen at the end , the first car just annihilates.
Also strange that the car cannot enter green a crossroad cars just after the crossroad are blocked - really they do so, creating mess :)

Which rules seem counter-intuitive?

Very elegant kata both in single- and multicar versions.
However, rules to pass green , especially at the end, seems counterintuitive for me
and demand additional handling, only to pass the tests

I feel your pain. But it kinda does -- the Description says that it uses x for rows and y for columns. It's messed up, but it is specified.

The description does not specify indexing order - x, y vs y,x.
Worse, the preloaded tests don't too.
So, if you initially chose the wrong order, you cannot understand what is problem with random tests.
Could you add such test, please, f.e. (is_efficient(11, 10, 5.5), True) ?

Could you clealry explain that only developer_id=1 must be checked,
or to drop records with develper_id = 2 from second random test?

I hadn't found this in the description,
so could not understand why my final test adds two extra records,
then I made ugly cheat hack to open solutions,
only to see - they all check developer_id=1 explicitly...

I would gladly disable gmp library for python but from previous experience this provokes a lot of nagging.

I may agree it is 7 kyu for Python,
but for JS, you need to write a prime check and search how to handle BigInt,
this still demands some work...

This passes all tests, but goes into endless recursion when arguments are not integer.
I think we should extend tests like it is done in Ruby version

The description is vague, I expected an integer arithmetics which is intuitive for "wallpaper planning",
but the correct solution demands most trivial interpretation.
Also don't understand why solution for one zero dimension is zero - in my opinion an abstract room 2x0x1 still needs 4m of wallpapers

The tests are fine, your function's name is formatDuration instead of format_duration. Rename it so the tests find it.

Looks like there is a problem in python tests

it starts with:

import codewars_test as test
from solution import format_duration

but when I run I get:

Traceback (most recent call last):
File "/workspace/default/tests.py", line 2, in
from solution import format_duration
ImportError: cannot import name 'format_duration' from 'solution' (/workspace/default/solution.py)