Thanks Hans. No matter how much I zoomed/squinted didn’t see a name on the original screenshot (so assumed (bad me!) it was unchanged). That makes it make more sense 
Out of interest I looked at one of these “single node” streets which disappeared - Recreation Road and also Cherry Tree Road here has a similar issue.
On the right is the “official” representation of the border between East Hampshire (on left side with the single node) and Waverley where the road is really in.
It is a bit of moot point as to whether any of Recreation Road is really in East Hampshire as clearly all the houses in it are in Waverley and the boundary “really” seems to stop at the notional “white line” junction with School Road. Of course OSM is more “schematic” and the node for the intersection is in both streets. The boundary as depicted in OSM seems to give the impression that a tiny bit of Recreation Road “really” is in East Hampshire. Ditto for Cherry Tree Road which ends where the aptly named Boundary Road starts.
I suppose if any part of Recreation Road and/or Cherry Tree Road is in East Hampshire would need to add another node to make a tiny way consisting of 2 nodes.
Further if I ask Google which county this two streets are in and tells me just Surrey (of which Waverley is a part of).
It would thus appear that by accident or design this change has “correctly” removed both streets from really being at all in East Hampshire.
Not quite sure how in OSM you can draw a road and a boundary so that the boundary is on the far side of the road but does not extend into any adjoining roads? In other OSM is seemingly giving a slightly false impression.
I don’t think overlaying the boundary with the road exactly in OSM is prudent either as such things get very messy. I might check some of the other 12 “lost” streets of East Hampshire to see if they basically show a similar situation.
I’ve been thinking about the issue where a street physically extends into a city but only registers a single node within the border, causing it to be excluded under the current “no single-node streets” rule.
What if the system treated the start node and end node of each linestring as valid nodes when evaluating whether a street segment belongs in a city? In other words:
• If a linestring crosses into a city, count both the entry point and exit point (or the first and last nodes inside the border) as part of that street’s node set.
• This would allow legitimate fragments of streets that do exist inside a city—just with minimal geometry—to be recognized as streets to complete.
• Streets that only touch the border from the outside would still be excluded, because they would truly only have one node in the city.
This could resolve the issue of valid border-crossing streets being lost, while still preventing the “single random node” artifacts the new data source created.
Here is a tricky one
Way: Prunella Place (389904040) | OpenStreetMap
This Prunella Place is a very short street so just has 2 nodes. However one is in Winchester and the other in Havant.
I just synched Havant and it got deleted.
Assume when I synch Winchester it will get removed there as well and so it is a street you effectively can’t “do” in CS any more!
If I add a “middle” node on the way for the street that also intersects with the boundary line in OSM, will that do the trick here? It is a bit fiddly in OSM to do that but may as well give it a go.
I have tried elsewhere adding a point both sides of the boundary line to see what effect it has but waiting for the city synch queue to settle down!
I’ve alternative 2, one node on each side of the boundary a couple of times, seems to work fine