I'm not really sure that this test is particularly memory-intensive, surely this is just an infrastructure issue rather than something in the test?

Killed process 23444, UID 501, (ruby) total-vm:1999508kB, anon-rss:550892kB, file-rss:1276kB

is the OOM message.

It keeps happening after this test and the test creates 50 reports so it could help to reduce the number of reports created
I'm not sure if this is the only problem but I've seen this a couple of times since the test was merged

I think you're right, it probably is this test. Maybe each report contains a lot of messages too?

I notice the time taken is quite long:

ReportTest#test_0011_should expire reports created 1 week ago = 15.72 s = .

And although I've increased swap on our slaves to avoid it, I notice most remaining SIGKILLs are in the vicinity of this test.