Memory leak exposed by exponential carbon starvation mortality

[glemieux]

Describe the issue

This issue is a mirror of ESCOMP/CTSM#4174. This test was added to cover existing mortality options recently with ctsm5.4.049.

It was noted in the CTSM SE meeting that this is may not be specific to this mortality mode since it is simply changing the way cmort is scaled from the parameter value.

Relevant log output

FAIL ERP_D_Ld90.f45_f45_mg37.I2000Clm60Fates.izumi_nag.clm-FatesColdCStarvationExp MEMLEAK memleak detected, memory went from 1156.350000 to 1392.020000 in 0 days

FATES tag

sci.1.92.7_api.46.0.0

Host land model tag

ctsm5.4.049

Machine

derecho

Other supported machine name

No response

Additional context

No response

[glemieux]

Asking claude (via cborg) for it's take it presented the following:

Given no missing-deallocate was found, the far more likely mechanism is that the exp model's sustained, never-fully-off mortality drives substantially more cohort create/terminate/fuse cycles — each cycle doing several small heap allocations for the PRT object's nested arrays — within the 90-day window...
To confirm rather than infer: compare cohort termination/recruitment counts between the two testmods (FATES already tracks term_nindivs_canopy/term_nindivs_ustory)