Thanks again Ernst for your information. I'll read the info you posted with interest and report back. Also thanks to David for his info.
This afternoon I checked and the swap file was 1 file of 100MB and 2mins old, so all the temp files that were created yesterday must have been genuine and correct and the 4.9GB peak I reached yesterday was all genuine valid data so it seems couldn't have been cleared earlier to free space.
Interestingly for the past few hours the hvs-2 swap usage has only peaked at around 500MB, presumably as only one Sentinal satellite happens to be downloading at a time.
For Bas and HVS-1 I do write the Tellicast received files to the ramdisk too, and wrote a Windows program to transfer them to two directory structures on my 2TB EUMETCast storage disk every 30 secs, one a copy of the received directories for use by Satpy or EumetCastView, while the other directory structure is the one needed by Satsignal software with some directories renamed or combined. This is so Satsignal can delete them when processed. I did try not deleting them so all programs could use the one directory structure but Satsignal processing took much longer as it kept searching through thousands of files for new ones. This only used 100MB of ramdisk in total so was no problem.
This won't work for HVS-2 as completed files are so much larger, so are written direct to the main EUMETCast storage disk. For the moment I may try risking using a 5.5GB ramdisk for hvs-2 tmp files too, and see how long it survives for. :)
I'm only just starting trying out Satpy/trollImage and Python so am not very good at it yet. I'm used to C and C++, so hopefully will get the hang of it quickly.