[tor-relays] risk of relay exploit

Hello,

Can anyone comment on the pros and cons of running a separate client-only 'tor' instance for local browsing while maintaining an independent instance configure as a relay?

My thought is that if high-resource adversary exploits/subverts the relay code and inserts a trojan/parasite, they would be in a position to monitor unencrypted traffic on the SOCKS port.

Is the probability of exploit significantly reduced for instances that do not accept relay connections? If so then keeping local browsing in a non-relay client would seem to be more secure since I understand that remote relays see only encrypted traffic.

Thanks