>People who find this a shocking and objectionable sign of bugs are generally people who have not build software at such large scales.

This leaves such a poor taste in my mouth. Perhaps the proper UX is to then _not display comment counts_ if your performance/cost tradeoff has determined that you can't display comment counts accurately. For higher comment counts it might be fine where a user isn't expected to read all 5000 replies(ignoring the edge case where its 5000 private/blocked accounts replying), but if a tweet has 2 replies then the user who clicks it expects to see those.

Other large platforms have been able to solve this issue either in UX or implementation, so "web scale" isn't a good excuse.

A tweet you can't see still exists. I think it's perfectly reasonable to display accurate counts even if it's based on information you don't have access to.

The click-thru page could then display a clarifying “+ x more tweets that are private or not available at this time.”

Yeah, sure, and kill the platform's greatest feature by MILES, which is the comment to like to RT ratio.

Is this the "ratio" I hear people talk about?

I understand the idea but it feels a little too online for my taste. I'm probably not the target audience. Just feels like someone decided the tea leaves falling a certain way MUST indicate something.

It's not always the case that a high reply/like ratio is someone being "owned", but it's obviously more concrete than tea leaves. Twitter's lack of a real downvote/dislike incentivizes people to reply to a bad post without leaving a like, and in my experience it's a pretty good metric. (The main exception is when a tweet is a prompt that intentionally asks for people to reply.)

Also I feel like I should add that "ratio" is a confusing term, because it can refer to the above example `reply-count / likes`, but can also be when a reply gets more likes than the tweet it's replying to: `reply-likes / OP-likes`.

> Is this the "ratio" I hear people talk about?

Yes.

Personally, I don't think "solving" minor inconsistency by eliminating a feature people like is the best approach. And from the way you talk about it, I gather you're not much of a Twitter user, so maybe you should give some deference to the people more familiar with the problem to decide whether this is a good choice or not?

If you have proof that other platforms have solved this problem at scale, I would be very interested to see it. Fundamentally, those totals are never going to be perfectly correct because a) people will be adding and removing tweets continuously, and b) even if continuously updating the numbers were worth the resources, people would hate having the numbers changing frequently.

It seems like this could easily be solved with "Some replies are hidden. [Settings](https://...)"

I remember when Twitter worked like this as well. It doesn't any more. When I see a tweet with a dozen replies and only one is visible, even in a private tab, it's very unlikely that eleven out of twelve are from locked accounts. It's even less likely that there wouldn't be at least one visible reply to one of those replies. I see this sort of thing often enough that I have no doubt that twitter is futzing with the visibility both of certain posters and of interactions with certain posters.