Thank you for your fast response. Sorry that I was not detailed enough.
With 'important method' I meant all the methods that are required for the kata.
Extra methods I created myself I am not logging, they are not part of the interface used by the tests.
I have some difficulty with providing more detailed information because I cannot see the test code itself. Only the output that it gives. But maybe I am unaware of a way to see the testcode.
I think it is test case 15.

describe block: "Mutating, adding, add_chaining and valence numbers consistencies"
it block: "Check all possible mutations and correct behavior of 'add' on the mutated atom"
There are no names given for the molecules.
The return values from the formula method with the 3 sequences I described are 'CH4', 'CBr4', and 'H'. As I would expect.

I cannot get testcase 27 to pass with Ruby. How can I figure out what test assertion is failing?
The error message is very generic: 'Testing raw formula with mutation'
The title of the testcase is: "Check all possible mutations and correct behavior of 'add' on the mutated atom"

I printed to the log whenever an important method was called, when I try that sequence myself all results are what I would expect.
The log I created:

"brancher call: [1]"
"mutate call: [[1, 1, "C"]]"
"closer call:"
"formula call:"

"brancher call: [1]"
"mutate call: [[1, 1, "C"]]"
"add call: [[1, 1, "Br"], [1, 1, "Br"], [1, 1, "Br"], [1, 1, "Br"]]"
"closer call:"
"formula call:"

"brancher call: [1]"
"mutate call: [[1, 1, "H"]]"
"closer call:"
"formula call:"```

"-> nope, that list of attacks leads actually to the numbers 2,2,1,1, boats 1 & 2 sunken, the "3" untouched."

I do indeed get the numbers 2,2,1,1 and thus 1 boat is sunken, 1 damaged, and 1 untouched. Just like the second sample test case is expecting. I am solving in ruby.

So this is in contrast to the result that Blind4Basics gave that 2 boats are sunken. I got confused by that comment and somehow overlooked that boat 1 had a length of 3 in the example given.

Anyway nice kata, I did get confused but that is not necessarily bad practice in my opinion.

I wrote here that my solution returned 11 in your example but 1064 in the medium pyramid but that was not true.
Actually I changed the test case myself to 1064 and when I fixed my algorithm the testcase failed.
Rookie mistake.

You are missing a couple of possibilities. For example the grouping 91, 74, 85, 81 gives 331.