Sure, a T/R bridge will do as well. But switching cables for different S-measurements and assuming this will not introduce considerable errors especially at the higher frequency end (workmanship dpendent) is probably a best wish only, especially if you want to do calibrated work (with properly calibrated cal kits). At least I dont trust me here. I have seen test sets for 500usds, the 75 ohms ones sometimes go even lower, a 50/75 ohms pad will help (loosing some dynamic range), I still prefer this over cable changing (also being lazy I admit).
Re. source failure, this in most cases seems to be the last stage amplifier die, which is on a ceramic substrate (at least on the older versions up to C). It can be replaced by a MMIC, there is some description if you google for it on an Agilent/Keysight page IIRC, but it is a little painfull. On a 8753B I had, it suddenly died (so as far as I am concerned, of the handfull of 8753 I had, the reliability statistics are bad). I got it to work again, but it was not fully stable at the very low end, output power limitations. This was due to RF choke issues required for the MMIC I used. It worked well from about 0.5 MHz onwards though. So it can be done if need be, considering a working source is probably almost as expensive as a VNA. Others doing that fix may have been more sucessfull here on the low end, not sure.