Extending De Gea and signing competition for him makes no sense if we are attempting to transform our playing style by introducing a goalkeeper who is comfortable being the extra man in deep build-up and a competent passer. We would regress every time De Gea is playing. Look at City’s backup that played in the FA Cup final, he’s a (less competent) version of Ederson.
Not only that, if he did get supplanted as the starter, we’d be stuck with De Gea as no one else would pay him a fraction of what we’re mooted to be offering him.
Having backups that don’t match the style and capabilities of your starters is a recipe for inconsistency and occasional disaster - see us without Casemiro or Bruno.
On goalkeeping errors, they tend to be very memorable as they often result in a goal being conceded, unlike errors in other positions and areas of the pitch. If you’ve only seen a goalkeeper in action a couple of times and they have blinder (think Friedel every time he played against us)/stinker you’re likely to be heavily influenced by this, even though it may not be reflective of their general performances, this is where statistical analysis can be helpful.
On the other hand, statistics that paint an incomplete picture have their own pitfalls. This analysis shows cross claims as the only metric to assess a keeper’s command of area. In reality, failed cross claims (like Costa’s in the World Cup) are often disastrous and result in at least a high quality chance if not a goal. Team style also plays into this. Defending set pieces with a higher line gives the goalkeeper more chances to claim a cross compared to a team that defends free-kicks deeper. That said, this can be influenced by the goalkeeper and their willingness to come off their line.
Whilst this analysis has its limitations given the availability of free data, it does mostly line up with my perception of the goalkeepers that I’ve watched a fair amount of.