GeezerVison

gamemode/core/libs/thirdparty/sh_deferred.lua (from brianhang/deferred-lua) never settles a chain when a step hands back a promise that is already rejected, e.g. return deferred.reject(err) from a :next() handler. Nothing downstream runs, no error is shown, and sometimes a misleading "Unhandled rejection" is printed even though a handler is attached.

deferred.resolve(1)
    :next(function() return deferred.reject("boom") end)
    :catch(function(err) print("caught", err) end) -- never runs

The adoption code copies the state first and then calls self:reject(), which does nothing once the state isn't pending. This PR applies the same four-line fix as upstream: brianhang/deferred-lua#1. That PR has the explanation, new tests, and a 3,000-chain differential check against the current version.

NutScript's own code doesn't return deferred.reject from handlers, so existing behaviour is unchanged. The only behaviour that changes is chains that used to hang. Schemas and plugins that ran into this can drop their workarounds. (We worked around it with a promise that rejects on the next tick.)

Checked on a GMod server with lua_openscript. With the current file, the repro chain return deferred.reject("boom") never settles, an "Unhandled rejection" error is printed even though a handler is attached, and the suite reports 56/60. With the patched file, everything settles and all 60 tests pass. The full output is in the upstream PR, and the tests and in-game check are at https://github.com/Michael-Tot/deferred-lua-tests.