The main consideration here is that a database with filestream can't be mirrored (and a mirrored database cannot have filestream added)
For 40kb average for the blob, filestream's not the best. The point at which filestream becomes more efficient is about 1MB files.
Can't really recommend HA/DR without seeing RTO and RPO figures, but I do quite like mirroring.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability