I think the fifteen COCA instances of "softwares" that Peter Taylor notes in his answer are worth looking at more closely.
Notably, five of the matches for "softwares" are from the same (Dutch) online source—and four of those are from a single article (published in 2017) on 3D software. This is unmistakable evidence that the usage is intentional. Moreover, the English proficiency of the author (LAB Wilson) seems to be quite high, although the style is regrettably wooden:
Indeed, a significant and recognized challenge for the beginner is the task of gaining a working knowledge of many 3D softwares and, in the case of applying geometric morphometric methods, the theoretical framework underpinning the comparison of shapes. Some of the additional benefits, such as interface with computational techniques to test form-function hypotheses, also come with the requirement for considerable specialist knowledge, e.g. engineering principles. 3D softwares vary significantly in cost, functionality, and ease-of-use, resulting in the need for a user to be familiar with several programs. A 'one for all' option is lacking, and most protocols involve the need to mix and match between softwares, demanding knowledge of file types and compatibility.
Clearly, for good or for ill, this author thinks that softwares is preferable to "software programs."
The author of the other article on the Dutch site (also published in 2017) shows a somewhat weaker grasp of English forms:
The development of multimedia services increases the appearing frequency of doctored images related to the political, advertising and personal attacks. The ease in availability of the photo editing softwares and high quality modern cameras has led to the manipulation of the image information to a great extent. Therefore, maintaining the authenticity of digital images is one of the primary concerns.
Four other COCA matches are from a single article (published in 2013) on bioinformatics analysis by a team of Chinese researchers. Again, there can be no doubt that the usage is intentional, but there is room to doubt the English fluency of the authors or their (human or automated) translator:
We used the PicTar (Krek et al. 2005) and Target Scan (Lewis et al. 2003) softwares to identify potential miR-184 target genes. To improve the accuracy of the analysis, we chose all the genes in predicted results of two softwares to use for GO term enrichment (http:// www.geneontology.org). If the P value less than 0.01 and at least 10 associated genes are in the GO term, it is considered these genes are enriched in the GO term.
Two other matches are from a single article published in 2016 in a periodical called Computers in Biology and Medicine. This author (Imon Banerjee) exhibits a generally strong (but not flawless) command of English:
Performing a comprehensive semantic annotation for all medical datasets is beyond the human capacity due to its massive volume. However, the efficient combination of man and machine can improve the speed and efficiency of annotation, and can offer ultimate understanding and utilization of anatomical data. In other words, semantic annotation softwares which extract the implicit content of the input data, parse all available symbolic information about patient history, and take into account the formalized medical knowledge, are becoming more and more relevant in this context.
The preceding sources—four articles across three periodicals/sites—account for eleven of the fifteen instances of "softwares" in the COCA search results. Of the remaining four COCA listings, one (from an NPR Science call-in show in 1998) seems very likely to be a typo or speech error, since it appears in the context of the phrase "the financial accounting package softwares that's":
And again that's just the operating system—the hardware and the operating system, so that has nothing to do with the Quicken software or whatever, the financial accounting package softwares that's running on top of it.
A second match from this final group of four solitary matches—this one published in 1995—may involve a punctuation (apostrophe placement) error:
[B]e aware that students tend to fit the hypotheses to their ability to manipulate the software, rather than to formulate hypotheses which necessitate their creative uses of the softwares' capabilities; ...
Transposing an apostrophe and an s is a common typographical error.
The remaining two instances, however, are more ambiguous. Here is one (published in in 1993) that could be read as a typo or as intentional:
Use of finite element analysis (FEA) and simulation softwares can determine the load needed to create 'first ply failure,' for complex-shaped designs.
The ambiguity arises because the author in this case might have been thinking, "I'm talking about FEA software and simulation software; those are two kinds of software or, more succinctly, two softwares."
And here is another (also published in 1993) that may not be a mere typo:
The most important issues in choosing fax softwares are ease of installation and adaptability to your modem. Sometimes it takes a few phone calls to tech support to get help with configuring the software to work with your specific fax modem.
Here, the author may have been so troubled by the wording "The most important issues in choosing fax software are ..." that he opted for softwares because it sounded better. Such a choice ignores the fact that are is there because the controlling noun in the sentence is issues, but I have seen many authors make the same false step in situations not involving software/softwares. It is possible, of course, that the author means to talk about choosing multiple fax software programs and considers softwares a legitimate word choice—but he seems pretty blithe about switching to "the software" (not "the softwares") in the next sentences. All in all, I think this instance is more likely than not to be a typo/braino error—but there is some possibility that the author believes in it.
Ultimately, the fifteen COCA matches for "softwares" break down as follows:
Eleven instances point to three definite sources of intentional use of "softwares" spread across four articles.
Two instances seem highly likely to involve accidental typographical errors.
One instance is (in my judgment) more likely than not to be a one-off error.
One instance is almost perfectly ambiguous, but may be reflect the author's view that "softwares" is a proper form.
That leaves us with four venues and five articles in which "softwares" probably appears intentionally—and notably, three of those articles come from authors who show a high level of English fluency.
Nevertheless, twelve or thirteen intentional matches in four or five articles is not as impressive as fifteen intentional matches in fifteen articles would have been. In fact, those low numbers are consistent with the view or hypothesis that use of "softwares" as a plural form is extremely rare among writers who are fluent in English. It is also noteworthy that the publication dates of the four articles in which the intentionality of the usage is clearest are relatively recent: 2013, 2016, 2017, and 2017. These may be signs of life in a post-copy-editor world.