What about MaxUnavailable?

[paulgmiller]

Surging helps minavailable if there's a single pod stuck unready but with low maxunavailable (1% for example) once you have a single unready replica you could be stuck. For example a 2 replica deployment with maxunavailable 1% is blocked with one unready pod but until you surge it to 100 pods you're still stuck.

Some options.

Mutate pdbs from maxunavailable to minavailble.
Dynamically increase maxunavailable when we surge and then restore it (hard for percent)
Increase not by the max surge but by whatever number is needed
Delete unready pods after we surge in pods if they become ready.
Some tradeoffs to consider. Need some data on how often people set low maxunavailable.

[paulgmiller]

Now that we're creating pods I am more open to mutating maxunavailable if custoemr opt in.