On my laptop with 16GB ram, the following command gets killed by the OOM killer. I believe quickbench is using too much memory from strictly reading the output of hledger register (about 700MB when piped to a file). I also notice quickbench is using considerable CPU time itself, which might skew timing results.

$ cd hledger
$ quickbench -f bench100k.sh 
Running 3 tests 1 times at 2025-02-27 13:15:14 CET:
Killed
$ echo $?
137

$ quickbench --help | head -n1
quickbench 1.0

$ hledger --version
hledger 1.30.1, linux-x86_64

Cause

I believe the cause is the use of readProcessWithExitCode. From the docs for readProcess:

readProcess forks an external process, reads its standard output strictly, blocking until the process terminates, and returns the output string.

Merge request incoming.