checksequenceverify – What prevents an arbitrarily high nSequence value when spending an OP_CSV output?

I have gone through the many related questions here on stack exchange and obviously read BIP68 and BIP112, but I still cannot seem to come to a clear answer for the following question: what is preventing a transaction that spends an output encumbered by OP_CSV from setting an arbitrarily high nSequence value?

Let’s imagine the following scenario:

• txA has a single output encumbered by 10 OP_CSV, meaning this output can only be included in a transaction that is mined 10 blocks after txA is confirmed
• txB spends a single input, which is the one output created from txA

The function that performs the check is

template 
bool GenericTransactionSignatureChecker::CheckSequence(const CScriptNum& nSequence) const
{
    // Relative lock times are supported by comparing the passed
    // in operand to the sequence number of the input.
    const int64_t txToSequence = (int64_t)txTo->vin[nIn].nSequence;

    ...
    omitted logic
    ...

    // Now that we know we're comparing apples-to-apples, the
    // comparison is a simple numeric one.
    if (nSequenceMasked > txToSequenceMasked)
        return false;

    return true;
}

Why should the spending transaction not be able to set nSequence for the input in such a way that txToSequenceMasked is always higher than nSequenceMasked? What exactly is preventing this? I feel like I’m missing something obvious here.

Hopefully my question makes sense, thanks!