Re your second question, I would say that perl has a tendency to do exactly that. Not only in allowing one to inline code from other languages, but in adopting facilities from other languages as it develops between versions; overloading and extension by modules etc.

Take for example PDL. It's still recognisably perl, but perl extended in a particular direction, adopting functionality from Matlab.

The community developed/supported nature of perl means that, if someone sees a need for a particular facility, they will very often write it as a module. Class::HPLOO for example allows one to create OO classes in a manner rather more like java. DBD::LDAP offers a degree of SQL syntax interaction with LDAP.

Perl has, IMHO, the biggest range of functionality of any programming language. A one liner can sometimes perform tasks that would require many lines in e.g. a shell script, and at the opposite end of the scale a fully functional GUI app can be written in perl. The fact that perl is so easy to use for quick and dirty fixup scripts perhaps leads people to underestimate the power and flexibility that it is capable of.

Update: As far as heavy use goes, I use perl for neural nets, which involves a very large number of FP calculations. It's not as quick as C - it's reputed to be 20-30% slower. But it's a lot easier to dig around in the guts of the code, and runtime compilation means I can modify the code, and not have to wait half an hour for the app to build before using it.