One thing that every manual/automated verification relies on is the consistency of ZFC + few inaccessible cardinals. I mean while humans have carefully thought about the flaws in that for century and didn't find any, it's not grounded in any clear reasoning why it is consistent.
Not at all saying it is inconsistent, but I personally find the liberal use of introducing inaccessible cardinals axioms a bit concerning.
Most natural problems like RH, twin primes, etc dont fundamentally rely on the inaccessible cardinals. Any problem that depends on them, there is much debate whether they can be called natural.
Many problems may not require even full zfc. zf, peano or even weaker may be enough.
There are also many theorem provers not just lean. lean depends on type theory, but metamath for example depends on zfc directly.
It may be a good but likely tedious exercise to convince yourself by translating the current proof to a zfc dependent prover.
> relies on is the consistency of ZFC + few inaccessible cardinals
I don't think so. They depend on some kind of type theory and formulate everything from those primitives.
It's possible that the question of consistency is too abstract to judge its significance. Why?
(1) A lot of the mathematics that was done in the 18th century was
obviously inconsistent and yet still useful.
(2) Trying to verify logical consistency of a theory is basically impossible
because of Goedel's 2nd Incompleteness Theorem.
Hitting a wall like this makes me think we're asking the wrong question.
Much of mathematics relies on these axioms. There is nothing special about the dependency of proof checkers or verification of this particular hypothesis. What is the purpose of this statement in this context?