Comment 8 for bug 1829813

> The regression potential field for the bug is a bit 'poor', but looking at the change itself I guess there's not much regression potential anyway. It seems safe.

Maybe we can mitigate by thinking about how to test this. Could you arrange testing with metadata of different sizes (<1Mb, 1Mb<x<10Mb, >10Mb) for example, to make sure that edge cases behave as expected? If that doesn't make sense that's fine, but please explain why.

To avoid holding this up I'll accept now, but please give some thought to what could regression and add that to your SRU verification please.