Logcat does contain:

10-29 19:12:16.827  6666  6666 E SplitCompat: Error installing additional splits
10-29 19:12:16.827  6666  6666 E SplitCompat: java.io.IOException: Unable to create directory: /data/user/0/org.chromium.chrome/files/splitcompat

which may be relevant.

Bisect points to https://chromium-review.googlesource.com/c/1304487 as the culprit, will revert.

There are a few linker-related entries in logcat:
10-29 19:12:17.102  6666  6700 W linker  : "/data/app/org.chromium.chrome-f7ohGipczS2cKdSH2YymHQ==/base.apk!/lib/arm64-v8a/crazy.libchromefortest.so" unused DT entry: type 0x6fffe000 arg 0x4c18
10-29 19:12:17.103  6666  6700 W linker  : "/data/app/org.chromium.chrome-f7ohGipczS2cKdSH2YymHQ==/base.apk!/lib/arm64-v8a/crazy.libchromefortest.so" unused DT entry: type 0x6fffe001 arg 0x12e58
10-29 19:12:17.103  6666  6700 W linker  : "/data/app/org.chromium.chrome-f7ohGipczS2cKdSH2YymHQ==/base.apk!/lib/arm64-v8a/crazy.libchromefortest.so" unused DT entry: type 0x6fffe003 arg 0x8

No idea how it went from digit@ to someone else...

Thank you. Note that the offending patch was reverted.

It appears the VR tests were loading native libraries that contain RELR relocations, with the system linker. These are only supported by the system linker starting with Android P, or by the Chromium linker, which is never used on N+.

Reverting the patch removed all RELR relocations from all native libraries.
It looks like a proper way to deal with that is to only compile libchrome.so with them, but not libchromefortest.so (which is used to load both libchrome.so or libmonochrome.so apparently).

Will look into it.

Also, relevant bug for RELR relocations: https://crbug.com/895194