And it was bad at other use cases, having used both (3270 and vt) you can do 3270 workloads on a vt(granted less efficiently) but vt style asynchronous updates were very hard on a 3270.

If I had to pick one I would pick the vt.

And footnote, while technically I have worked with 3270's, my experience was using rexx on vse/cms. That is to say my use of the 3270 was hampered by about 5 levels of application framework, perhaps if I used them more directly I would be more charitable.

Right, the great advantage of the 3270-style block-mode model was that your 0.5-MIPS mainframe didn't have to handle an interrupt for every keystroke, so it could serve, I don't know, 100 terminals. Same as HTML <form> and half-duplex line-mode terminals with local echo.

An in-between version is how csh and vi would set eol2 (I think?) to ^[ so that most user interaction could be done in "cooked mode" until you hit ESC or ^D. This still required an interrupt for every keystroke on your 1-MIPS PDP-11, but at least it didn't require a context switch to the shell or editor process in order to echo normal printing characters. The kernel would handle echoing characters back to the terminal, deleting characters with ^? or ^H, and deleting words with ^W.

(As I understand it. I used csh and vi, but never on a PDP-11.)