Ok so I have now successfully repaired the S-760 crystal problem. Since I couldn't find anything about this issue online - or that much about S-760 repairs in general (anyone still using the '760??) I thought I'd post my findings here in case anyone else runs into this problem (I am fairly convinced that the damage to the crystals was caused by a glue that Roland coated them in, so others having this problem might not be as unlikely as sounds!
).
Short version:
These three crystals (X5/X6/X8 on the main board) have a direct effect over two vital things in the S-760 - sampling frequencies and overall sound output. If/when they die you will either lose sampling at various frequencies and/or you will lose audio output altogether. The originals components are hard to find, particularly the 16.934 MHz, but you can replace them with HC49 crystals of the same value. I used crystals that matched the data sheet of the originals as closely as possible - including the load capacitance - and hopefully, everything now seems to be working great.
Longer version if you're interested:
The original crystals are Epsom CA-301 series crystals. I couldn't find anywhere that had all three values in stock. Digikey and Mouser each sell the 12.288 MHz and the 18.432 MHz respectively but with minimum delivery charges you're looking at paying around £13 for each crystal. I couldn't find anywhere that still has the 16.9344 MHz in stock so, since I was going to have to substitute that one anyway, I thought I may as well try substituting them all.
I read online that the load capacitance of the crystals is important in some oscillator circuits but not in others, depending on whether they are in parallel or series. Since my understanding of basic electronics is not good enough to tell the difference by looking at a schematic, I figured if I could match the load capacitance of the originals that would be a good belt 'n braces approach. All three originals have a load capacitance of 18pF so that's what I looked for in my replacements, as well as matching the frequency values.
I am still none the wiser as to whether load capacitance matters in this case but for the benefit of others that can tell the difference, here is the schematic for the three crystal oscillator circuits and the associated IC - note that the attendant caps and resistors are all SMD's mounted to the solder side of the main board underneath the crystals.
The HC49 package is a similar size to the originals but a completely different orientation, so a bit of fiddling is required to get them to fit. There are a lot of exposed test points on the S-760 main board, so I put insulating tube on the folded over legs of the HC49's - a bit of insulating tape on the board underneath the crystals might have been a good idea too, but I managed without.
First one in:
All in:
Hope this info might help someone out - thanks to Paul for initial advice!