this was added due to model limitations, just like plan mode
“models are too fast and do toolcalls only when talking! lets restrict that!” we said with plan mode
but now that models have gotten better at using toolcalls only when needed, plan mode makes no sense
same thing with queue:
“models dont know how to handle different prompts at the same time and stop doing the ‘long task’”
“lets fix this by making so messages are sent after they're done”
this is mostly fixed with frontier models, and next models wont even flinch at this
queue should be removed from every harness just like plan mode is.