@hiredman is this also true of your patches?

should I look at them, or not look at them?

oh, look at them, the first patch in the series changes the rawcode stuff to only reconstruct the environment that the expression actually needs

to be the clear, the changes in those patches seem to have removed the oome from the example on the ticket, and I don't get the oome from the one other linked ticket I tried either

ok, will do

in testing it was a little tricky, since triggering the oome with a finite range size seemed hit or miss

yeah, it can be fiddly

inspecting the macro expansion of the into-chan definition with the patches applied, it doesn't store coll in the state vector any more, and ch and close? get nil'ed after the last use (but that doesn't effect the oome and is as the loop is exiting)