I've been gardening my way through these posts manually since this vote's direction became clear (down to less than 130 pages remaining from over 400!), and while doing so have made a number of observations from which effective conclusions can be drawn.
Firstly, discounting outlier posts where the sisters tag was applicable (largely involving cosplay) or single-post examples (usually a user adding the tag while uploading about a week ago, when the sisters tag was common enough to be suggested), the effective totality of the sisters tag has been added to Kancolle posts by only three users: user #1153489, user #455356, and user #492049. Of these, the earliest edits found were performed by user #1153489 in June and July of 2025, and were only performed on select images of the Kongou class, seemingly done as additions to a larger project by that user to replace detached sleeves with shoulder cutout on images of said class. This means that, despite the Kantai Collection tag existing for over a decade, there were no posts where the sisters tag had been added earlier than last year.
Secondly, the vast majority of the tag was added by user #455356 beginning in April of this year, with user #492049 contributing to the tag beginning at the end of April. Both users continuing to add it until the creation of this thread on May 9th, meaning the vast majority of the 400 pages of posts tagged Kantai Collection and Sisters had the latter tag added in the space of about a month. There are no similarities in situations by which these posts were added: instead, they followed scenarios and ratings across the board, linked only by the characters involved: the Kongous, the Yamatos, the Nagatos, the Fusos, DesDiv 6, and the entirety of pool #10686. Given the lack of similarity many of these characters exhibit, and the lack of interaction on many of these posts, one can only conclude that the sisters tag was added by these two users en masse based solely on the character tags involved, which sends us into a larger but separate argument of what exactly is supposed to be the point of the sisters tag is at all. This also echoes the concern raised by Hillside Moose that the determination of "sister" in a Kancolle or other personification context relies more on metaknowledge of the objects being personified than it does any objective measure of siblinghood, which again raises the spectre of tag pollution.
Returning back to argument at hand, however, the concentrated and extremely recent nature of these tag edits serves to refute NNT's points earlier in this thread regarding consensus: the larger community of Kantai Collection uppers and taggers, it seems, do not consider shipgirls of the same class to be familial relations, regardless of the valid question of what familial relations of personifications are supposed to look like. Whether the edits by user #1153489 last year comprise the very first time a familial relationship was tagged on a Kancolle post, or simply the oldest example to have been unreverted, is a question that may be interesting in the sociological sense but is irrelevant to this discussion. Also irrelevant is the argument over the larger question of familial tags in general, or the arguments raised by Evvk over how similar personification games such as Azur Lane treat familial relations.
In the end, it seems that the Kancolle community has largely been holding true to the wiki instructions to not tag shipgirls as sisters, and the recent spate of tagging thereof was an aberration from the norm rather than an adjustment of it.
In conclusion, so as to preserve the sisters tag where properly applied on cosplay posts, I'll reject the opening BUR and then finish manually gardening the remaining offenders over the next few days.