just tried to recreate the problem you're having, but everything seems
fine on my end (4.5.7 also). i loaded a 1000ms soundfile into [buffer~
1000], and scrubbed my way outside the buffer window, into a second
monitor up to 36000ms and then back and nothing crashed. the audio
scrubbed fine too when the mouse went back over it.
hope this helps you a bit, let us know what the problem was if you find it.
Yes (kind of) I used to get this quite often prior to 4.5.6 especially with high load granular patches - i never used the waveform mousing though just trying to read longer than buffer length with play~ did it for me. The problem was hard to reproduce, and would crash max sometimes but not every time. I ended up using lots of clip~ objects to constrain any play values that were likely to go out of range.
I don't recall having that problem recently though - did something get fixed? Or did i just become more thorough when working out the maths?!?