Still crashes for me in Devel::Peek::Dump - but at least I can see what you have got, which will help.
If I can stop writing responses (both here and elsewhere) I might even find time to do some thinking.

My bug report is here.
So far only one response (from Tony Cook):

It's a bug in the 32-bit x86 ABI, see:
https://rt.perl.org/Ticket/Display.html?id=125710
where we first saw the problem, and:
http://stackoverflow.com/questions/22816095/signalling-nan-was-corrupted-when-returning-from-x86-function-flds-fstps-of-x87
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57484#c12
Material which that perlmonks discussion appears to cover.

Pretty much what Buk has been saying (surprise, surprise), though I haven't yet fully digested those threads.

I've followed up Tony's reply by pointing out the very similar behaviour I see with MSWin32-x64-multi-thread-ld architectures - as I'm not so sure *that* behaviour can be explained in terms of the 32-bit x86, x87 ABI.
No repsonses to that as yet.

Cheers,
Rob