Following from the EMPF HelpLine in Indianapolis, prepared by Jeff Kukelhan. 
This supplements previous replys to this subject.

62/36/2 paste has been used in the hybrid industry for years to solder to Ag 
bearing thickfilm materials (it cuts down on leaching).  Apparently it does 
not cause any problems in hybrid applications.  Just remember, in order for 
electromigration to occur, there has to be condensed water present.  Keep 
condensed water off of the finished assembly, and there will be no 
migration.  Remember that Sn and Cu also will migrate just as easily as 
silver under moist conditions.
 ----------

From: TechNet-request
Subject: Re: Soldering to Gold
Date:  12 Dec 95 9:32PM
From: Karen Tellefsen <[log in to unmask]>
  
 ----------------------------------------------------------------------------

>I have a question about reflow soldering to gold with 2% silver paste.
>
>I heard some  folk lore at our company about using 2% silver solder paste
>on gold plated lands. The legend states that 62/36/2 paste on gold has
>long term reliability implications, these joints being subject to
>dendritic growth to a greater degree than 63/37 solder to gold.
>
>If anyone is aware of published data on this subject, I would be very
>grateful if you would give me the references so I may review the
>information.
>
>Also, if anyone is using silver solder on gold at this time, I would be
>interested in hearing about your experience.
>
I think this has more to do with Ag in the solder than the Au lands.
Ag is known to be a bad electromigrator because of the low activation
energy for Ag dissolution/redeposition reactions and the soluability of
Ag-organic acid salts.  Krumbien's review on electromigration discusses
this in general, though it doesn't discuss 62/36/2 specifically.

However, we at Alpha Metals frequently run electromigration tests for
pastes with 62/36/2 as well 63/37 powder to ensure reliability with
both alloys.