Without the use of pars, is there any way to spot check that there are no biases or errors in your projected figure making?

Not so sure that is clear, so I'll give one example that I know you already take into account.

2 yos, 3 yos, and other lightly raced horses tend to slowly improve over time. Without that insight, you might have a tendency to project figures for them that are slightly slower than the horses are actually earning when making your variant. Over an extended period of time, you could then wind up with overall figures that slowly shrink enough to not be comparable to figures produced 6 months earlier - even though they would be accurate in their current snapshot in time to each other.

I suppose there could be other biases and/or errors like that.

Some figure makers spot check all their figures against the PARS every couple of months to make sure they aren't shrinking or accelerating due to figure making errors.

At the same time, I agree 100% that the PARS themselves can and probably do change over the years because of the breed and other reasons.

I have always found this inter-generational figure issue quite interesting and perplexing.