FWIW I recently encountered this and bisected it to:

e36a9c459d7d1644f401eb49f3684c6fe42f7948 is the first bad commit
commit e36a9c459d7d1644f401eb49f3684c6fe42f7948
Author: Alexandre Julliard <julliard@winehq.org>
Date:   Tue May 5 11:54:11 2020 +0200

    kernel32: Move volume functions to kernelbase.
    
    Signed-off-by: Alexandre Julliard <julliard@winehq.org>

 dlls/kernel32/kernel32.spec     |   40 +-
 dlls/kernel32/volume.c          | 1675 +-----------------------------------
 dlls/kernelbase/Makefile.in     |    3 +-
 dlls/kernelbase/kernelbase.spec |   40 +-
 dlls/kernelbase/volume.c        | 1777 +++++++++++++++++++++++++++++++++++++++
 5 files changed, 1851 insertions(+), 1684 deletions(-)

--

-Austin
GPG: 267B CC1F 053F 0749 (expires 2021/02/18)

Yeah, it is pretty much possible, those function added to kernelbase could be the last bit after which the size exceeded the threshold on your build.

I thought of fixing the problem by removing a dozen of random functions from kernelbase, but I did not test this solution yet.