I believe you would use the single WC card as the WC Input for MADI card #1 - and tell that card to sync to WC.  Then, use the internal jumper (if PCIe/HDSPe series MADI card) and feed the 1st card to the 2nd card internally.  Set the 2nd card to sync over "Sync In" (or whatever it is called).  Should do the trick.

With regard to "(2)- How does fibre optic I/O use WC?"  MADI's integrated clock is pretty poor, but the RME MADI cards all have Steadyclock to recover a WC from a jittery MADI signal with less than 2ns of jitter (Steadyclock was developed specifically for MADI IIRC).  That being the case, you could still sync Card #1 to your MADI input (Steadyclock will spit out a low-jitter clock), and still feed the internal Sync from Card #1 to Card #2 (as above, but w/o the additional WC card)

For #1 Yes, but these internal ports only work on the HDSPe PCIe versions of the card (these ports aren't active on the HDSP/PCI version IIRC).  The "Sync In/Out" connections are on the main MADI card (not the WC card).  I believe they use a PC CD-Player type connection for patching (type of cable that used to go from the CD-Player to the soundcard in old PC's) - but don't quote me on that.

For #2 - I was suggesting that you still sync Card #1 with the MADI input itself, and then use the internal jumper to feed card #1's clock over to card #2.  Might help, might not depending on what's going on at the other end.

Feeding a BNC Wordclock from the main/master MADI device to RME MADI Card #1 would likely be your best bet IMO.  Then slave Card #1 to card #2 with the internal sync connections as mentioned above.  But if you don't have any room for the extra WC card, then syncing over MADI *should* work (as #2 mentioned).

Are the 2 MADI devices you are connecting to the two RME Cards sync'd together?  What are the MADI devices involved, and how are they currently sync'd to each other?  If those 2 devices aren't sync'd, then that is the first thing you'll need to address...