Currently the network has a very inconsistent track record when it comes to injecting thousands separators in numbers >= 4 digits. Sometimes on the same page there will be commas in one element and not in another.
The primary advantage to using thousands separators is that it makes a number easier to read, at least for me.
The primary disadvantage is that it makes them harder to read for folks who use different languages or locales, where commas are actually decimals, and thousands separators may be decimals, spaces or other characters (or not even an existent concept).
I've seen arguments that the SE network should use comma separators because "the site is American" or because "English is the language of the Internet." I am not quite obnoxious enough to say I agree with either of those statements. And as a Canadian living in the United States, where arguments about m/d/y vs. d/m/y, gallons vs. liters and miles vs. kilometers have plagued me my entire adult life, it would be fantastic to have some consensus about this and to see some consistency implemented across the network.
I understand that localization efforts are in progress. But I still think it's a valuable discussion to have - should we force thousands separators, prevent them, or cater to the individual locale? The downsides to catering per language are (a) screen shots and other artifacts may still seem strange to users of different locales and (b) what if my computer is set to US-English but I would prefer not to see thousands separators? Perhaps the answer is to always use the shorthand (e.g. 5.4K, 137K) instead of listing out the entire number in some places but not others - then the discussion about separators becomes moot.
PLEASE NOTE: This is not a bug report about the individual cases where the use of thousands separators is inconsistent. There are plenty of those already, and if pluralization has taught us anything, it seems to be an unwelcome way of reporting the issues. I hope it is the intention for the localization efforts to push all number and other common formatting tasks through common functions to eliminate this chance for discrepancy.
This post is an attempt to discuss what the policy should be, in general, not to discuss individual instances of this inconsistency.