Firstly, let me introduce myself:
I am of the age that I bought a brand new ZX-Spectrum with my hard-earned first pocket money back in the 1980s.
Countless hours of playing games and writing BASIC programs followed.
After about 40 years I found this ZX-Spectrum in the attic and decided to make it work again. My intention is to keep it as original as possible.
My Speccy is an Issue3 - 48K. Only IC1 (ULA) is socket-ed.
So far, I have executed the following works:
- replaced the keyboard membrane & put an insulating tape over the RF-convertor,
- recapped the axial capacitors (btw: the old ones measured still OK)
- performed the Composite mod. (with the capacitor)
- after the Composite mod, I replaced TR1 & TR2 with BC549C's (rotated) in order to enhance the contrast and brightness
- measured resistance/continuities and the various power-voltages, which are all fine
- tested the ZX-Spectrum functionally OK.
Now all is well, apart from a strange video problem. I've done some research but couldn't find this particular issue let alone the solution.
The problem is as follows:
- when I turn on the Spectrum it works fine. All functions work well.
- after a while (varies between 30s to up to 15min) there are stutters and flashes on the screen
- and a little (seconds to minutes) later the image 'dims' ; the background and border gradually turn black (within 5 - 15 seconds) until it is completely black. And if I let it go even longer, eventually the video-convertor gives up and stops giving video signal.
- strangely enough, occasionally it suddenly recovers to its normal video state, but eventually it always ends up in the 'black screen' state.
- in the mean time, the Spectrum remains fully functional, even after the screen is black
- After the Spectrum has been powered-down (sometimes more than an hour), it has recovered to normal operation
I have done quite some measurements with a multimeter and an oscilloscoop. In the 'dark state' all signals appear to be OK as far as I have measured, apart from one: the signal on the base of TR2 - originating from IC14 - pin 13 appears to be the guilty party: in the 'dark' state this signal is a (almost) flat 4.25V signal, while in the normal working state it gives a nice waveform.
The attached screenshot shows the signal at the base of TR1 (yellow) and the base of TR2 (red), in the 'black screen situation'.
This leads me to suspect that IC14 may be defective.
However, since I do not yet have a replacement and prefer to do as little soldering as possible, my question is if my diagnoses appears to be OK? Or may there be other possible causes and if so, which other tests/measurements could I perform in order to check these?
Many thanks in advance for your thoughts...
Standby alert 
