I bought a cheap Russian stopwatch on eBay about 10 years ago, and its accuracy has been wanting.
I finally removed the back and visually inspected it.
It is a bit agricultural, but lets press on.
Above, the hairspring hints the cause of the accuracy issues.
The end curve of the hairspring between the end collet (upper left) and regulator arm and through to the dog leg formed in the spring is not a constant radius from the jewel, and further, the dog leg is badly formed and distorts the hairspring coils off the balance pin centre. If you compare the space between the outer spring turns at the bottom of the pic and top of the pic, the off centre distortion is evident.
When the watch stabilises and amplitude maximises, the inner part of the dog leg section actually touches the adjacent coil, which of course will compromise rate.
The measured beat error (a measure of oscillator asymmetry) was 7ms, huge, and probably the main contribution to fairly poor positional error, and observed rate change with main spring wind.
A couple of minutes with tweezers to carefully reshape the hairspring in situ gave a much more pleasing shape.
The hairspring end curve was fairly easy to reshape, so it probably isn't steel. Nevertheless, the watch was demagnetised as a precaution.
Since for timing longer intervals, the stopwatch is almost entirely used flat on its back and with a nearly fully wound spring, the rate was adjusted in that condition. As it turned out, beat error was less than 0.3ms as a result of the spring reshaping and there was no need to tweak the isochronism adjustment
Above, a pic of the timegrapher. The watch runs like this hour in hour out, and when bumped, it stabilises without the glitch caused by the touching turns.
The stopwatch was replaced with a Casio MS-70W which has been an excellent tool, and apart from the rubber ring that holds the strap on, has required no maintenance. So, whilst the mechanical stopwatch is not ‘needed', it now works well and is an acceptable reserve that does not require a battery.