What, if any, version tags should be applied to a Q&A that has answers relevant to a range of versions?
This is another situation that supports the original version tag guidelines as used on other exchanges - don't use one unless the question or issue is specifically related to the version. That said, this question doesn't appear to be considering initial tagging, but rather editing, modification, and merging after the fact. In such a case, and with the horses already out of the barn...
If you're already making an edit and you know it's not a version specific question (ie timeless), particularly if your changes are related to cleaning up duplicate questions from differing versions, remove the version tags.
If not consolidating questions of differing versions (but still doing an edit - tag edits just to add or remove a version tag should be avoided in my opinion, again because of trivial edits and differing philosophies on version tag use), I would probably leave version tags alone because of the current situation with them (being that they are frequently added to all new questions).
I do not support adding either the highest applicable version tag or especially a pair of tags that are supposed to indicate the range of version applicable. Slots available or not, that seems like it goes against a couple of different tagging system principals and further dilutes the value of version tags (from either viewpoint).
I think version tags reach their maximum potential when they help those answering questions for a particular product to be able to identify whether the question is relevant to the version(s) of the software that they have installed. This assumes more importance for commercial software where some upgrades require an additional purchase or ongoing maintenance to be paid, than it does for the no cost upgrades of open source software.
When deciding whether, and what, version tags should be applied to a "timeless" question, my thought process derives from the commercial ArcGIS rather than the open source QGIS which is being used in the particular "timeless" Q&A that @underdark asked after.
Current consensus on QGIS version tagging, gauged from Double tags qgis and qgis-2.0, suggests to me that the QGIS sub-community would, on "timeless" Q&As currently favour using a single qgis tag and dispensing with any QGIS version tags. When tagged this way the multi-version answers of a "timeless" Q&A remain open-ended in both directions so to @underdark's comment "Should the qgis-2.6 tag be removed from the Q?" I would say "Yes, according to what appears to be consensus on using version tags with QGIS questions."
However, if this were a timeless ArcGIS (typically Desktop) Q&A, then my answer would incorporate that a large proportion of ArcGIS Desktop questions are tagged with the version that is being used in the question, and I think that protocol can be easily expanded to accommodate "timeless" Q&As. Instead of QGIS 2.6 and 3.0, I will use, as an example, an equivalent "timeless" Q&A that involves ArcGIS Desktop 10.3 and 11.0.
As always, whether any version tags are used on ArcGIS Desktop questions will depend on whether there are any "spare" tag slots available, because I do not advocate replacing any more important tags with version tags (which I consider to be of important but noticeably lesser value).
If answers on a "timeless" Q&A range from ArcGIS Desktop 10.3 to 11.0 then:
- if 5 important tags were already being used, then I would place no version tags
- if only 4 important tags were already being used, then I would add only the highest numbered version tag, in this case arcgis-11.0 so that it can act to tag the current vintage of the Q&A
- if 3 or less important tags were already being used, then I would add both the lowest and highest numbered version tags, in this case arcgis-10.3 andarcgis-11.0 (and, if an earlier ArcGIS 10.0 Q&A were later merged into this one, then I would version tag it arcgis-10.0 and arcgis-11.0 instead)