They always have checks and balances, and you are quite right in wondering about it.
Take the usual life cycle of a young financial analyst in one of these firms (not even doing quant stuff)
He makes model, it is sent to his superior who checks it and tells him the out put is wrong, the PE ratios he has calculated for the firm are through the roof, at something liek 18x while the rest of the industry is always at around 3x. (numbers are fictional).
Check your numbers.
Further, the lowly analyst has also made a million errors in even reaching this stage - from incorrect formatting, to incorrect formulas, incorrect linking, to the dreaded ref error.
They do this stuff constantly, and the ones who can do it in their sleep and not die of exhaustion to the brutal hours are the ones who get through to the next level.
People have their fingers burnt regularly, and every day. Bad models still manage to slip through, but are more often than not, caught, or handled.
So people in the industry are frightened witless that something like a bad model may escape into being published.
It took several sequences of failure for the whale to amass the positions he did, and it is an uncommon sequence. Which is why its such a failure on the industry's part.
You've hit on exactly why mistakes in excel spreadsheets don't get corrected. Evveryone checks the output, many people check the results of intermediate calculations, but in most cases people don't check the code if those are ok.
Everything goes well until either the data changes enough to make the error in the code matter, or some other code is added which relies on assumptions about how the broken code works, and the sheet starts pumping out garbage.