The question whether a Golden Age tango recording is in the public domain in the United States seems to depend on a particular historical question to which I don't know the answer (but I would love to, so please help if you can!): how much time did it pass between first distribution of the recording in Argentina and the first distribution (or at least the first offering for distribution) of the recording in the United States? If for a recording the answer is "30 days or less" then it is very likely that the recording is in the public domain in the US.
Let me explain for those interested in the intricacies of copyright (while stressing that none of what is said below should be construed as legal advice). Until the advent of the current US law, the Copyright Act of 1976, copyright in the US was regulated by The Copyright Act of 1909. According to the 1909 Act copyright lasted for 28 years after publication unless it was renewed in its last year for a second copyright term. To quote from Circular 15a of the United States Copyright Office:
Under the 1909 act, federal copyright was secured on the date a work was published or, for unpublished works, on the date of registration. A copyright lasted for a first term of 28 years from the date it was secured. The copyright was eligible for renewal during the final, that is, 28th year, of the first term. If renewed, the copyright was extended for a second, or renewal, term of 28 years. If it was not renewed, the copyright expired at the end of the first 28-year term, and the work is no longer protected by copyright.If copyright on the work was not renewed in its last year then the work entered to the public domain. The 1976 Copyright Act substantially (and retroactively) increased the length of the 28 year renewal term however it did not change the copyright status of those works that already entered to the public domain due to the lack of request for renewal.
A request for renewal was supposed to have been submitted to the Copyright Office and all submitted renewal requests were recorded in the Catalog of Copyright Entries. Thanks to the hard work of government transparency activists all of these Catalogs have been scanned, text-recognized, and uploaded to the Internet Archive: https://archive.org/details/copyrightrecords . If there is no record for the renewal request among the Catalog entries 28 years after the first publication of a work, copyright expired on it in the end of the 28 year period and the work entered to public domain.
I've checked around a dozen better known Golden Age Argentine tango recordings in the Catalog and it appears that renewal requests have not been submitted for them at the appropriate time; on this basis I believe the same is probably the case with the overwhelming majority of Argentine tango recordings. Take for example the vals Pobre flor by the orchestra of Alfredo de Angelis. It was recorded and published in 1946 by Odeon Records. Suppose its publication date was two months after its recording date, on the 09/01/1946; then copyright on it in the US lasted until 09/01/1974 unless a renewal request was submitted sometime between 09/01/1973 and 09/01/1974. I have checked the Catalog of Copyright Entries for 1973, 1974, 1975 and found no renewal requests for the song (or for any other works of Alfredo de Angelis for that matter). Sans the renewal request Pobre flor entered to public domain in the US 28 years after its publication and stayed there at least until January 1, 1996.
Why January 1, 1996? Well, because with that effective date the Uruguay Round Agreements Act of 1994 retroactively restored copyright to certain foreign works whose US copyright protection had been lost because of noncompliance with formalities of US law, in particular noncompliance with failure to renew. Foreign works that met certain criteria automatically got treated as if their copyright holders had submitted a renewal request, and since the renewal terms got extended to ridiculously long periods (typically the full term became the last author's death plus 70 years), eligible Golden Age tango recordings are still most likely within their extended copyright term.
Is this the end of story? Not quite. The Uruguay Round Agreements Act only applied to works that met four criteria (see Circular 38b of the United States Copyright Office). The first three of these are almost certainly met by Golden Age Argentine tangos, however the last criteria is curious:
4. If published, the work must have been first published in an eligible country and not published in the United States during the 30-day period following its first publication in the eligible country.
Now the US copyright law defines "publication" as:
the distribution of copies or phonorecords of a work to the public by sale or other transfer of ownership, or by rental, lease, or lending. The offering to distribute copies or phonorecords to a group of people for purposes of further distribution, public performance, or public display, constitutes publication.
In other words, if in our example Odeon Records published Pobre flor in Argentina in 09/01/1946, but did only so much as to offer it to a US entity for the purpose of distribution between 09/01/1946 and 09/30/1946, then Pobre flor does not meet the criteria set out by the Uruguay Round Agreements Act, and hence it is still in the public domain in the US!
To me it sounds quite plausible that at least an offer for sale in the US was made within a short amount of time given the already global nature of music distribution (i.e. the originally German Odeon Records became part, through multiple acquisitions and merger, of the London based EMI in 1931 which had a well established distribution network in the US). I do not know the answer for sure, however, and I have no access to historical documents describing the publication practices of recording companies at the time to verify it beyond doubt. 30 days does seem to be a short period of time, so we would really need some expert looking into this issue. Apart from expert knowledge about publishing practices other forms of evidence would also be interesting, i.e. does someone know of periodicals of the time that mentioned new tango releases (and offered them for sale) that are known to have been distributed in the US? (I.e. "Caras y Caretas" or similar publications could be consulted - but were they available in the US in a timely manner? (Thanks Bernhardt for the suggestion.))
If Pobre flor were in the public domain in the US, one would be free to use one's own shellac-to-mp3 transfer of it under a Youtube video without needing to worry about copyright claims made by The Orchard Music or anyone else, Youtube being a US incorporated company. (This would be so despite the song potentially being protected by copyright in countries other than the US, for instance in Argentina; Youtube might then be compelled to impose viewing restrictions on the video in Argentina, but not in the US. I'm emphasizing "own transfer" because the transfer from shellac itself can create a copyright for the person who makes the transfer depending on how much added value is created in the course of the transfer, resulting in this aspect of the newly created mp3 file falling under copyright protection.)
Long story short: I'm looking for historians' input here who are familiar with the marketing and distribution practices of Odeon, RCA etc. and could comment on the issue of time passing between distribution in Argentina and (offerings of) distribution in the US. If you know someone (who might know someone) who could know these things and could back it up with credible historical evidence, please put us in touch!
If I find additional relevant information on the topic not mentioned in the comments below, I will update the post accordingly.