T20030_test1j(normal) unexpected failure

Primarily, this test fails in the thread_sanitizer job, along with a bunch of others (#22520). However, it has failed once (in my data) in a "normal" validate job: https://gitlab.haskell.org/ghc/ghc/-/jobs/1322105, aarch64-linux-deb10-validate.

Example:

=====> 1300 of 9257 [0, 0, 0]
=====> 1400 of 9257 [0, 0, 0]
Compile failed (exit code 139) errors were:
[ 1 of 13] Compiling A[boot]          ( A.hs-boot, A.o-boot )
Segmentation fault (core dumped)
*** unexpected failure for T20030_test1j(normal)
=====> 1500 of 9257 [0, 1, 0]

All cases since last April:

create temp table jt as
select rowid from job_trace where trace match '"unexpected failure for T20030_test1j"'
;
create temp table jj as
select job_id, json ->> '$.name' like '%thread_sanitizer' is_thread_sanitizer
from job
;
select json ->> '$.created_at', jj.job_id
from jj
left join jt
on jj.job_id = jt.rowid
left join job on
jj.job_id = job.job_id
where 
is_thread_sanitizer = 0
and jt.rowid is not null
order by is_thread_sanitizer

---------

json ->> '$.created_at'   job_id 
------------------------  -------
2022-06-16T10:43:55.428Z  1084081
2022-06-16T13:43:27.707Z  1084214
2022-09-28T12:30:20.615Z  1184977
2022-09-28T12:30:21.831Z  1184980
2022-09-28T12:30:21.908Z  1184982
2022-09-28T12:30:22.697Z  1185002
2023-01-17T19:45:40.552Z  1310940
2023-01-25T13:32:27.839Z  1322105

Edited Feb 08, 2023 by Bryan R

To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information