This was going to be a question but now it has turned into a heads up for anyone who encounters the same issue.
While trying to install ESXi 5.1 I would regularly get as far as "Select a disk to install or upgrade" but when I tried to choose the SD card as the storage device the installer would crash. It would also crash if I tried to get the details of the SD card (F1).
The following backtraces (all with slight variations) would then scroll continuously up the screen until I power cycled the server:
Installing to the SAS drives would proceed without a complaint so I was pretty sure the install disk wasn't corrupted.
The relevent lines of the backtrace (as far as I could tell) leading up to the final assert were:
/lib/python2.6/site-packages/partedmodule.so [0x1ffd21ff] (usually between 7 & 9 on the stack but on one it was as high as 11)
/usr/lib/libparted-1.8.so.0(ped_disk_new+0xaa) [0x10012b2a] (always one higher than the previous line on the stack)
/usr/lib/libparted-1.8.so.0(ped_disk_add_partition+0x2cf) [0x10011d9f] (consistently at call 15 on the stack)
/usr/lib/libparted-1.8.so.0(ped_assert+0x3b) [0x10009f2b]
My next thought was a dodgy SD card but it mounted fine on my Windows 7 box.
I googled the assert and found a report of a bug in partedUtil in ESXi 5.0 (http://lists.gnu.org/archive/html/bug-parted/2012-07/msg00001.html). It seemed to be something to do with the partition not falling neatly on cylinder boundaries or something so I repartioned the SD card on a Mac (I tried in Windows but I didn't seem to be able to modify the partitions from within Disk Management nor could I delete any of the non-primary drives in the extended partition?!).
After re-partitioning (to 1 partition formatted as FAT) I could install ESXi 5.1 fine. I assume it was the same libparted bug I referenced above but it took me a while to track it down so hopefully this post helps someone else out there who bumps into the same issue.