Creating a tag for one or two - or even a handful of questions, regardless of whether the new tag would be "mutually exclusive" or "a completely different thing" vs. the most applicable existing tag, is usually counter-productive.
A similar discussion has already taken place on this meta, about the game tag.
- For askers, they connect experts with questions they will be able to answer - they allow you to direct your question to the people most
able to answer it.
- For answerers, they are for sorting your questions into specific, well-defined categories. You can follow tags, get updates on changes,
even mails and other notifications.
- Tags can be used to report on broad statistics, tracking activity over time, etc.
- Tags can be used for awarding badges.
Feel free to propose a new tag for things, but be mindful that tag fragmentation is worse than not enough tags.
How does this relate to pascal, free-pascal, delphi and friends?
As you may or may not be aware, the pascal "tag family" isn't exactly thriving. You mentioned 4 tags, and together they don't even really come close to meet the "badgeable status" threshold of 100 questions.
Is fragmentation a mandatory criterium, more important than "tag mutually exclusive descriptions"?
I'd say yes. Take vba for example. It has 231 opened, non-deleted questions... and I'm not even close to start considering splitting it into excel-vba and access-vba, even though the two object models have pretty much nothing in common. Fragmenting a small tag does more harm than anything else. C# 6.0 introduces features that don't compile under earlier versions - yet there's no c#-6.0 tag, and when C# reviewers see C# 6.0 features they know what they're seeing. java-8 was hastily introduced, and ended up synonymized with java -- and Java is the biggest tag on this site. Tag fragmentation is a serious problem.
There is a minimum acceptable fragmentation for a tag?
I'm happy you came to meta, but would have preferred you came to meta before creating the new tags. Tags are managed by the entire community, and their creation/synonymization/burnination are always discussed on meta. The "create new tags" privilege doesn't grant one carte blanche for creating whichever tags they want unilaterally. Add to that, that Code Review is currently in a semi-graduated state where privilege thresholds are way below what they should be (e.g. create tags at 150 vs. 1500), and that this situation has been dragging for quite a while, and you'll get a glimpse of the frustration regulars have been experiencing for two years now (just take a look at meta posts tagged beta-progress). So no, there's no "carved in stone" minimum acceptable threshold to fragment a tag - only these things require community consensus. Fragmenting a marginal tag into even smaller tags that apply to only a handful of questions, helps no one.
Don't get me wrong, a lot of Code Review regulars are complete pedants. We love being pedantic. We love to have everything in its place, and using the right names for the right things... when the time is right.
IMHO now is still too early to consider breaking up pascal into more specialized tags; I'd go as far as recommend folding back delphi-xe into delphi, and free-pascal into pascal (i.e. remove the synonym and retag the few questions involved to get rid of the minor tags).