@gevs @rb2 thanks for elaborating 
Anyhow, I see some issues with this approach. Please correct me if I’m wrong.
1.) The package is checked against an arbitrary transaction in the blockchain. If an attacker is able to replace both the package (.tgz) and the signature file (.tgz.sig), and he includes an appropriate transaction in the blockchain, then the check would also succeed. I.e. the binaries are checked for integrity, but the are not checked for the issuers identity.
This could e.g. be exploited if someone claims to operate a mirror for the binaries and replaces the files by malicious ones at some arbitrary point in time.
2.) Both .tgz and .tgz are published via http, which makes it easy for a man-in-the middle to replace the content of both files.
3.) The sha256 is determined by querying another nem node (igalice3.nem.ninja in this case) for the transaction content via http. The user is forced to trust the node and anyone between the two (potential man-in-the-middle) to not tamper with the content.
So, as I see it atm, the integrity + identity check via PGP signature has been replaced by an integrity check only.
I would like to see the PGP signature file be added for download again.
Regards,
Kc
P.S.: Maybe it’s also possible to include an identity check against the issuer of the transaction and query the users own xem node instead a remote one.