The spec requirement at https://html.spec.whatwg.org/multipage/input.html#number-state-(type=number):valid-floating-point-number-4 just says:

If the value of the element is not a valid floating-point number, then set it to the empty string instead.

But it doesn’t seem like a strict implementation of the “valid floating-point number” checking algorithm can on its own determine whether the value represents a number less than or equal to Number.MAX_VALUE.

    // parseDouble() allows the string to start with a '+' or to end with a '.' but those
    // are not valid floating point numbers as per HTML.
    if (characters.front() == '+' || characters.back() == '.')
        return std::nullopt;
    size_t parsedLength = 0;
    double number = parseDouble(characters, parsedLength);
    return parsedLength == characters.size() && std::isfinite(number) ? number : std::optional<double>();

My implementation for Ladybird looks like this:

bool is_valid_floating_point_number (StringView string)
{
    GenericLexer lexer { string };
    // 1. Optionally, a U+002D HYPHEN-MINUS character (-).
    lexer.consume_specific('-');
    // 2. One or both of the following, in the given order:
    // 2.1. A series of one or more ASCII digits.
    bool has_leading_digits = !lexer.consume_while(is_ascii_digit).is_empty();
    // 2.2. Both of the following, in the given order:
    // 2.2.1. A single U+002E FULL STOP character (.).
    if (lexer.consume_specific('.')) {
        // 2.2.2. A series of one or more ASCII digits.
        if (lexer.consume_while(is_ascii_digit).is_empty())
            return false;
    } else if (!has_leading_digits) {
        // Doesn’t begin with digits, doesn’t begin with a full stop followed by digits.
        return false;
    }
    // 3. Optionally:
    // 3.1. Either a U+0065 LATIN SMALL LETTER E character (e) or a U+0045 LATIN CAPITAL
    //      LETTER E character (E).
    if (lexer.consume_specific('e') || lexer.consume_specific('E')) {
        // 3.2. Optionally, a U+002D HYPHEN-MINUS character (-) or U+002B PLUS SIGN
        //      character (+).
        lexer.consume_specific('-') || lexer.consume_specific('+');
        // 3.3. A series of one or more ASCII digits.
        if (lexer.consume_while(is_ascii_digit).is_empty())
            return false;
    }
    return lexer.tell_remaining() == 0;
}

In order to pass the WPT tests, I need to call that is_valid_floating_point_number function and additionally parse the value:

    } else if (type_state() == HTMLInputElement::TypeAttributeState::Number) {
        // https://html.spec.whatwg.org/multipage/input.html#number-state-(type=number):value-sanitization-algorithm
        // If the value of the element is not a valid floating-point number, then set it
        // to the empty string instead.
        if (!is_valid_floating_point_number(value))
            return String {};
        auto maybe_value = parse_floating_point_number(value);
        // AD-HOC: The spec doesn’t require these checks — but other engines do them, and
        // there’s a WPT case which tests that the value is less than Number.MAX_VALUE.
        if (!maybe_value.has_value() || !isfinite(maybe_value.value()))
            return String {};