maybe a typo on "JDK 25 G1/Parallel/Serial GC changes"

[muzinian]

Hi, Thomas.
In the post on "JDK 25 G1/Parallel/Serial GC changes"

tschatzl.github.io/_posts/2025-08-12-jdk25-g1-serial-parallel-gc-changes.md

Line 39 in 2dbc7b6

Determining the candidate regions is where the cause of the high pause times lies: evacuation efficiency is a weighted sum of amount of live data in the region and the size of the remembered set for that region; G1 prefers to evacuate regions with low liveness and small remembered sets as they are fastest to evacuate. However, at the time the candidate regions are selected, G1 does not have information about the remembered sets, actually G1 plans to recreate this information just after this selection. Until now, G1 solely relied on liveness, assuming that the remembered set size is proportional to remembered set size.

You mentioned at the end of this section

Until now, G1 solely relied on liveness, assuming that the remembered set size is proportional to remembered set size.
is this a typo?I suspect you meant to say the remembered set is proportional to the amount of liveness?If I’m mistaken, could you please explain what it means for the remembered set size is proportional to remembered set size?