Copied from a note in JDK-8282704:

I've attached the latest RSS monitor results with the fix for this
bug (JDK-8282704) and JDK-8282314 in place on my M1 MacMini:

do_all_check_stress_results.jdk-19+13_redo_1.log

My analysis scripts no longer complain about possible leaks.

However, I do still see something suspicious in the max_value values
for StressWrapper_suspendthrd003. Here's all the max_value values
from the most recent run:

$ grep max_value !$
grep max_value do_all_check_stress_results.jdk-19+13_redo_1.log
max_value=72128
max_value=179728
max_value=178400
max_value=142480
max_value=218560
max_value=208368
max_value=43600
max_value=131280
max_value=171328
max_value=202512
max_value=209456
max_value=180880
max_value=65776
max_value=180688
max_value=178720
max_value=69104
max_value=175888
max_value=179760
max_value=63856
max_value=173008
max_value=177680
max_value=125360
max_value=166176
max_value=171936
max_value=157424
max_value=213184
max_value=216320
max_value=62688
max_value=170752
max_value=174416
max_value=142448
max_value=193328
max_value=186352
max_value=56096
max_value=166816
max_value=172960
max_value=568272
max_value=3120272
max_value=1692416

The last three values are for StressWrapper_suspendthrd003 and they
are much higher than any of the other StressWrappers... Since we don't
see an OOME anymore, I have to guess that this is possibly some
native memory leak.

Here's the full entries for StressWrapper_suspendthrd003:

StressWrapper_suspendthrd003 release:
n_values=240
max_value=568272
min_value=55584
max_increase=8496
min_increase=7488
Value did not increase 1 times.
Value decreased 67 times.
Value increased 171 times

StressWrapper_suspendthrd003 fastdebug:
n_values=240
max_value=3120272
min_value=179136
max_increase=354256
min_increase=44256
Value decreased 76 times.
Value increased 163 times

StressWrapper_suspendthrd003 slowdebug:
n_values=240
max_value=1692416
min_value=182048
max_increase=170768
min_increase=15872
Value decreased 73 times.
Value increased 166 times

I haven't used NMT for a long time so I'll likely investigate
this possible issue with another bug...