I believe I may have finally found the cause of this issue…was a schema tangle.
Not all SQL includes a full schema.table syntax…It looks like some connections were being released back to pool defaulting to a different schema which then causes previously working non fully qualified queries to then fail when they reuse these connections. Since a"Select 1" will still pass validation these connections are not pruned creating the persistant intermittant behaviours observed.
After fixing the schema issue I now have a few days of logs without these errors.
It is probably a good idea to standardise on fully qualifed queries going forward since these make no assumptions on default schema (most queries are already but it appears there are exceptions)