Two witnesses can charge that one or both of a pair of other witnesses are thieves (or some other psul) and disqualify them from testifying. However, two witnesses would not be believed if they charge that the testimony of other witnesses is a lie. We would be left with a case of trei u’trei, a standoff. Tosfos (B"K 72b d"h ain) asks: why in this second case would we not believe the accusing witnesses that their counterpart’s testimony is a lie migu that they could just accuse their counterparts of being thieves and disqualify them?
The first answer of Tosfos focuses on metziyus. Migu assumes that were someone to lie, he would tell the best lie possible; a defect in a claim is evidence of its truthfulness. This sevara does not apply when speaking of two witnesses because two people practically would be unable to coordinate a story to fabricate a better lie.
The second answer of Tosfos is based on a din. We know that the testimony of 100 witnesses is no better proof than the testimony of 2 witnesses: trei k’me’ah. We also know that the testimony of witnesses is stronger proof than migu. So if 100 witnesses are no better proof than 2 witnesses, QED that 2 witnesses + a migu are no better proof than 2 witnesses alone.
According to the second answer of Tosfos, migu adds no credibility. However, according to the first answer of Tosfos, were it theoretically possible to apply the sevara of migu, it would make a difference. What is the point of debate?
R’ Elchanan (Koveitz Shiurim) writes that there are two ways to understand trei k’me’ah.
1) Relative: One type of proof can be weighed against a qualitatively different type, e.g. ruba v’chazakah ruba adif, rov is better proof than chazakah, but there is no difference between a greater or lesser quantity of the same type of proof. All proof through testimony, be it from 2 or 100 witnesses, amounts qualitatively to the same thing.
2) Objective: testimony of two witnesses is by definition the highest degree of proof possible and therefore cannot be trumped by anything, even a greater quantity of witnesses.
If trei k’me’ah just tells us that different quantities of the same proof equal out, then adding a qualitatively different type of proof to the mix, e.g. adding migu to one side of the scale, tips the balance. However, if trei k’me’ah means that by definition there is no proof of any type superior to that of two witnesses, adding migu to the scale has no effect.
Question/observation: according to R' Elchanan's first approach, the scales only move when we compare and weigh qualitatively different forms of proof against each other, not different quantities. Sounds to me like this means that sugyos that invoke two chazakos beating one (e.g. the discussion in Nidah 2-3) work only if the chazakos being weighed are qualitatively different. It would be interesting to test if we could really come up with qualitative differences between all the various forms of chazakah, but that type of project is beyond me at this point.