And if one party trends more towards violent extremism, they get boosted the same as before?

What's that got to do with accidental effects of unrelated software changes?

if your hate speech filter hits one party harder than the other, that doesn't necessarily mean it is broken

The point is to have these statistics remain the same for the individual groups in between changes, not to measure them against each other.

So then why make any changes?

Why not? If it doesn't matter but it's controversial there's no reason not to change it.

What if you're changing the spam filter?