On Tue, 18 Jun 2024 13:47:49 +0200
Alexander Graf [off-list ref] wrote:

IMHO the big fat disclaimer should be in the argument name. 
"reserve_mem" to me sounds like it actually guarantees a reservation - 
which it doesn't. Can we name it more along the lines of "debug" (to 
indicate it's not for production data) or "phoenix" (usually gets reborn 
out of ashes, but you can never know for sure): "debug_mem", / 
"phoenix_mem"?

I don't see any reason it will not reserve memory. That doesn't seem to
be the issue.  What is not guaranteed is that it will be in the same
location as last time. So the name is correct. It's not
"reserve_consistent_memory" ;-)

-- Steve