Not precisely what you're asking but I have, very occasionally, REPLed into remote clojure servers and rewritten running code live to fix time-critical bugs in production. But I'll admit i'm unsure whether this is an argument for or against Xtremely interactive development.
Darklang does offer something they call trace-based development where unhandled requests basically do initiate an interactive process to write handling code. I'm under the impression though that this is not intended for production time.
In general I'm loath to make any generalization about what kind of mistakes are good or bad and when. Luckily we have a language landscape which lets people make up their own minds.
Darklang's traces are not the same thing as a REPL. The user gets sent a normal 400 error, and the request trace is saved so that later, the developer can "replay" the request (on the live server, with full context).
While Dark does support the general idea of doing live code in production, it's not a Lisp dialect and is very much a functional language.
Not precisely what you're asking but I have, very occasionally, REPLed into remote clojure servers and rewritten running code live to fix time-critical bugs in production. But I'll admit i'm unsure whether this is an argument for or against Xtremely interactive development.
Darklang does offer something they call trace-based development where unhandled requests basically do initiate an interactive process to write handling code. I'm under the impression though that this is not intended for production time.
In general I'm loath to make any generalization about what kind of mistakes are good or bad and when. Luckily we have a language landscape which lets people make up their own minds.