I agree as well that 1 writesf~ is preferable.
Also, as you may or may not know, if you need *very long* audio recording support, you can try building the branch which includes the CAF file format which uses 64 bit indices:
Doing basic recording or playback of files > 4 GB should work, but accessing via sample position, etc won't work unless you also build Pd with double support otherwise Pd's internal float will clip your larger sample indices.
I'm hoping for this branch to be integrated in the next Pd version, so it would be great if more people can test it. It also includes a number of bug fixes, better format handling, etc.