As they say, even a stopped clock is still right twice a day.

A backwards clock is right 4 times a day.

A clock running at 10,000 rpm is right 14,398,560 times a day.

The relevant metric is surely the proportion of time for which the clockhands are within epsilon of the correct value.

Pretty sure this metric will be constant (provided the hands aren't perfectly synced). The frequency of confluence is proportional to the difference in rates, but the duration is inversely proportional.

https://www.desmos.com/calculator/45k0rrjwo0

So the better way would be something like average angular distance? If it's close to correct, but just off by a second or so - then that average will be very small.

Same problem. The distribution of misalignments will be uniform.

Picture a clock frozen at the 12 o'clock position. It will be exactly aligned at midnight and noon, and exactly misaligned at 6am and 6pm. The angular misalignment will smoothly go from 0 to 180 degrees, then back at the same rate. The time it takes to move into and out of alignment will change, but the average misalignment will always be 90 degrees (absolute value).

The only exception is clocks which proceed at the exact same rate. Here they will both be stuck in sync, locked into whatever the initial misalignment was.

Texas has had win after win, I've not seen too much policy out of there that I haven't liked recently.