The thermal research is purely academic FEA sims, some using a laughable 2W power. I'll remain a skeptic.
https://nazrulanuar.files.wordpress.com/2008/11/b26_s0060p171-1751.pdf"... Reducing the copper heatsink thickness to form a single gauge leadframe, a thinner version of the dual gauge leadframe, was perceived to present various electrical, thermal and reliability failure mode and mechanisms, specifically die cracking. This study presents an in-depth simulation and analysis to determine whether the thinner heatsink leadframe version is capable of achieving the defined customer product requirements on their electrical, thermal and reliability performances in a DPAK package."
" Results revealed that the safest leadframe thickness range is within 0.5 - 1.0mm. The original diepad thickness falls within the safe range. Selecting the 0.5mm as the new diepad thickness for the single gauge leadframe design was as straightforward as referring from the leads thickness of the original dual gauge design. Assembly, ambient electrical, and reliability testing were carried out for the carrier devices (non-auto and automotive types) in DPAK production line. Zero-hour electrical and reliability test results revealed that products satisfied all test specification requirements for both automotive and non-automotive devices."
https://ieeexplore.ieee.org/document/7365171If the thickness limit is only about die cracking, no mention of thermals- they might have forgotten higher power dissipation >20W?
For TO-220 single gauge, I find the thin tab warps and doesn't contact a heatsink well, becasue of the bolt shoulder/insulator bushing contact area. This is why I hate the sneaky cheap package.
I think JEDEC should have issued a new package part number, they did for all the old parts such as TO-3, TO-92 creating more chaos when it was totally unnecessary.