I'd also consider tweaking mine to a lower percentage if that helped, or possibly even a flat 0.038MB if we wanted an absolute guarantee that there was no conceivable way it could increase by more than 1MB in the space of a year.
I don't like fixed increases in particular either. Percentage based movements in both directions would be nice, but the primary problem with those is preventing the system from being gamed. Even with 10%, eventually this 10% is going to be a lot. Who's to say that at a later date, such movements would be technologically acceptable?
The thing to bear in mind is we'll never make any decision if we're too afraid to make a change because there's a possibility that it might need changing at a later date. Plus, the good news is, it would only require a soft fork to restrict it later. But yes, movements in both directions, increases and decreases alike would be ideal. This also helps as a disincentive to game the system with artificial transactions because your change would be undone next diff period if demand isn't genuine.
You could argue that it may already be quite late/near impossible to make such 'drastic' changes. I've been giving this some thought, but I'm not entirely sure. I'd like to see some combination of the following:
1) % changes either up or down.
2) Adjustments that either align with difficulty adjustments (not sure if this makes thing complicated or riskier, hence the latter) or monthly adjustments.
3) Fixed maximum cap. Since we can't predict what the state of the network and underlying technology/hardware will be far in the future, it is best to create a top-maximum cap a few years in the future. Yes, I know that this requires more changes later but it is better than nothing or 'risking'/hoping miners are honest, et. al.