Weird test mixup

I got a weird test failure while testing my forking refactor patches on 
Cirrus CI 
(https://cirrus-ci.com/task/5880724448870400?logs=test_running#L121):

> [16:52:39.753] Summary of Failures:
> [16:52:39.753] 
> [16:52:39.753] 66/73 postgresql:intarray-running / intarray-running/regress                       ERROR
6.27s  exit status 1
 
> [16:52:39.753] 
> [16:52:39.753] Ok:                 72  
> [16:52:39.753] Expected Fail:      0   
> [16:52:39.753] Fail:               1   
> [16:52:39.753] Unexpected Pass:    0   
> [16:52:39.753] Skipped:            0   
> [16:52:39.753] Timeout:            0   
> [16:52:39.753] 
> [16:52:39.753] Full log written to /tmp/cirrus-ci-build/build/meson-logs/testlog-running.txt

And:

> diff -U3 /tmp/cirrus-ci-build/contrib/intarray/expected/_int.out
/tmp/cirrus-ci-build/build/testrun/intarray-running/regress/results/_int.out
> --- /tmp/cirrus-ci-build/contrib/intarray/expected/_int.out    2024-03-14 16:48:48.690367000 +0000
> +++ /tmp/cirrus-ci-build/build/testrun/intarray-running/regress/results/_int.out    2024-03-14 16:52:05.759444000
+0000
> @@ -804,6 +804,7 @@
>  
>  DROP INDEX text_idx;
>  CREATE INDEX text_idx on test__int using gin ( a gin__int_ops );
> +ERROR:  error triggered for injection point gin-leave-leaf-split-incomplete
>  SELECT count(*) from test__int WHERE a && '{23,50}';
>   count 
>  -------
> @@ -877,6 +878,7 @@
>  (1 row)
>  
>  DROP INDEX text_idx;
> +ERROR:  index "text_idx" does not exist
>  -- Repeat the same queries with an extended data set. The data set is the
>  -- same that we used before, except that each element in the array is
>  -- repeated three times, offset by 1000 and 2000. For example, {1, 5}

Somehow the 'gin-leave-leaf-split-incomplete' injection point was active 
in the 'intarray' test. That makes no sense. That injection point is 
only used by the test in src/test/modules/gin/. Perhaps that ran at the 
same time as the intarray test? But they run in separate instances, with 
different data directories. And the 'gin' test passed.

I'm completely stumped. Anyone have a theory?

-- 
Heikki Linnakangas
Neon (https://neon.tech)