I know SIV doesn't address release of unverified plaintexts and iirc, AEZ and RIV (https://fse.rub.de/slides/fse_talk_lucks.pdf) do. Are there any robust AE schemes nearing standardisation / deployment / "mainstream" like SIV is?
Alas, there are no standardized robust schemes, with AEZ making it to the CAESAR final portfolio being the most likely scenario of that happening.
Point being, its inclusion in the final CAESAR portfolio is far from clear at this point.
Disclaimer: I'm not the GP.
To the best of my knowledge, that's why he designed AEZ under the encode-then-encipher paradigm - encode the plaintext with \tau trailing zero bits, then encipher it with (what is intended to be) a length-preserving SPRP.
Being an SPRP, the inverse is also a PRP, and thus any adulteration of the ciphertext results in a pseudorandom decryption, preventing RUP (and causing the trailing zeroes to fail to validate).
Constructing the SPRP then becomes the interesting part - a fascinating construction can be found in https://arxiv.org/abs/1310.4050 (which neither uses a block cipher entire, paying the cost of multiple invocations per block, _nor_ gives up on proofs - instead, it extends Cook's Elastic Cipher construction to arbitrary lengths, using the _round function_ directly, with security reductions to the original cipher the round function was taken from.)