That isn't that much of a fix, because you can add hole digging/filling in any amount to the schedule of any socially necessary worker. Someone who digs and fills a hole in the course of welding a pipeline will engage in more labor than someone who just welds the pipeline, but they will both accomplish the same goal of welding the pipeline.
You might say, "by labor I mean the difficulty of accomplishing the task when operating at the greatest possible efficiency," but an idealized, infinitely smart programmer could write perfect, bug-free code as fast as they could type, and an ideal physicist could make all possible theoretical inferences instantly.
I don't understand your counter-argument; time spent doing useless things on a schedule in which non-useless work is accomplished doesn't mean that the time doing useless things is suddenly valuable. Any capitalist knows that they don't want to be paying someone to do work that doesn't produce value. The person who only welds the pipeline spends less time getting the task done, which is obviously the aim here.
Socially necessary labour is defined as labour performed under normal ("averagae") conditions of production using tools of average production potential. At least in the Marxian schema (which I'm not really here to defend) skilled difficult labour is only a greater magnitude of simple, unskilled labour. For Marx, the quantitative relationship between them (and indeed between value and socially necessary labour time) was not only not necessary to derive, but arguably impossible in a complex economy. The important part was that it worked. Value works "behind the backs" of participants in an economy to shift their labour to more profitable regions of the economy.
You're right in that the further you get from "the labor theory of value" the better you do, like mixing a bowl of Doritos with a bowl of celery. Smith admitted a little bit of market valuation, and in doing so was a little bit better than completely wrong.
Quoting from your link,
>For the modern reader an alternative version of Smith’s calculation may
seem more natural: the ‘labour commanded’ by a commodity represents the time you would have to work (say,
at the average wage) in order to buy the commodity. In both cases – Smith’s version and the modern one –
the calculation of labour commanded is the price of the item divided by some measure of the wage, usually an
average.
Clearly, N hours of work at an average wage W is just a roundabout way of expressing a quantity of money NW, and from there Smith believed in market pricing.
Smith's error (which turned him to "market pricing") was not recognising the dual character of labour, which is solved in the Marxian schema; capitalists purchase labour-power, which is a commodity with an exchange value (i.e the labour commanded by it is the amount required to reproduce it) of W and a use value (value in use) of concrete labour itself, which is the capacity to create value. Unlike Smith's idea, this is compatible with both varying value of money, varying wages (assuming there is something to vary the wages) and useful in explaining the role of profit.
On the whole, though, I dispute the notion that the LTV must be abandoned; you are already starting from the position that a bowl full of celery is a good thing, but I'd say that we'd do just fine with the Doritos, and the case for a bowl of celery lacks the explanatory power of the bowl of Doritos.
You might say, "by labor I mean the difficulty of accomplishing the task when operating at the greatest possible efficiency," but an idealized, infinitely smart programmer could write perfect, bug-free code as fast as they could type, and an ideal physicist could make all possible theoretical inferences instantly.