I break very long URLs up with a percentage sign % which has the effect of putting a single return in the line. Seriously long addresses (like some of those on the IBM Lotus Smartsuite support site) may be broken up with two or more % splices.
As you say, it seems Opera is to blame. Despite my fondness for many of its features I no longer keep Opera on any of my systems, not even for site testing purposes. I play with Opera whenever a new release comes out, but that's as far as it goes.
I normally build in what I know are 'safe' features to the sites I work on. This comes with experience and over time you get a strong sense of what can and cannot work consistently over different browsers. I came to the reluctant conclusion some time ago that until many of the issues Opera seems to have are ironed out I can no longer spend any serious time using it for daily use.
I test unusual site designs in IE, NetScape and sometimes Firebird on Windows, and IE and Jaguar on the Mac. If the results are consistent, that's good enough for me. This is not meant to be critical of Opera's userbase; more of some of the fundamental issues it has with what should be simple code rendering.
Web browsers normally interpret the % sign in a web address as a single space; the equivalent of hitting your spacebar once, so they effectively ignore the % sign hence it is useful for breaking up URLs.