I’m still experiencing packet loss and I can see it occurring via packetlosstest.com, and also our neighbour is with EE via Openreach and experiencing the same issues as us.
How much is a "small amount"?
Packet loss of even 0.1% *can* badly affect TCP performance, when talking to a distant server. The maximum speed you can get from a TCP session is a function of the packet loss and round-trip time:
https://en.wikipedia.org/wiki/TCP_tuning#Packet_loss
e.g. for a connection with 0.1% packet loss and 70ms RTT (East coast USA):
1460 / 0.07 / sqrt(0.001) = 659560 bytes/sec
Multiply by 8 for bits per second and you get ~5.2 Mbps
Zen has been in touch and asked if we’d like to upgrade because the contract is up next month. I’m tempted to upgrade to the next package up, which is 500/70.
But by doing so, is that likely to result in more packet loss because the connection is faster? Or could it go the other way and result in more speed but less packet loss, or possibly even no packet loss altogether?
It will almost certainly make no difference to packet loss. It's just a change in profile on the OLT.
Low levels of packet loss are hard to measure directly, and you can get false readings due to (for example) the router itself dropping pings addressed to itself when it's busy.
IMO, the best approach is to do a single-threaded speed test, to several servers some distance away (e.g. Europe or USA). If you find you get lower speeds to servers with a larger round-trip time, then this is evidence of low-level packet loss and you can use the formula above to work backwards to estimate the amount.
If you are getting good speedtest performance to all these sites, or the pattern is inconsistent, then likely packet loss is just a red herring.